Acoustically measured postural stability with visual feedback

Material Information

Acoustically measured postural stability with visual feedback
Shuman, Dennis, 1946- ( Dissertant )
Doty, Keith L. ( Thesis advisor )
White, Keith D. ( Reviewer )
Berg, W. Keith ( Reviewer )
Childers, Donald G. ( Reviewer )
Dawson, William W. ( Reviewer )
Place of Publication:
Gainesville, Fla.
University of Florida
Publication Date:
Copyright Date:
Physical Description:
xii, 226 leaves : ill. ; 28 cm.


Subjects / Keywords:
Buffer storage ( jstor )
Bytes ( jstor )
Error rates ( jstor )
Microcomputers ( jstor )
Negative feedback ( jstor )
Retinal images ( jstor )
Signals ( jstor )
Software ( jstor )
Timing devices ( jstor )
Vibration ( jstor )
Dissertations, Academic -- Electrical Engineering -- UF
Electrical Engineering thesis Ph. D
Equilibrium (Physiology) ( lcsh )
Equilibrium (Physiology) -- Data processing ( lcsh )
bibliography ( marcgt )
non-fiction ( marcgt )
thesis ( marcgt )


Postural stability is a function of several weighted sensory inputs to a central neural controller. A new laboratory tool was developed to help quantify body sway performances in general and to explor the contribution of visual input as a feedback control signal in particular. By utilizing body sway data collected online for the control of a surrounding pattern’s movement, the visual feedback loop was altered to permit further characterization of postural control. A parallel processing microcomputer system measured body sway in any direction, an degenerated visual stimuli based on mathematical analyses of the incoming sway data. Position and orientation determinations were accomplished by ascertaining the transit times of sequential sound fronts from head attached acoustic sources to an array of fixed receivers surrounding the subject. With two sources employed, sampling rates of 25 Hz were achieved with position and orientation accuracies better than 1 mm and 0.2” within a volume sufficiently large to accommodate the extreme limits of body sway. The system managed the horizontal position of a vertical grid shadow cast upon a vertically oriented cylindrical screen encompassing the subject’s field of view. When governed by the sway data, the grid position response delay introduced was less than 90 msec, and the position accuracy was better than 0.5” over its full 32” range with unity feedback gain. A study of 33 subjects employing stimulus movements governed by body sway, with feedback gains ranging from -2 to +2, demonstrated differential influences on sway frequency spectra by absolute and relative (to those normally expected) retinal image motions. Only stimulus movements that were both spatially and temporally correlated with sway behaved as true feedback control. Sway spectra thus observed were ordered directly with positive feedback gains and inversely with negative feedback gains. Additionally, those spectra exhibited large peaks and valleys that displayed frequency consistency across feedback gain magnitudes but reversed with feedback gain polarity. The importance of regarding sway energy about 1 Hz was confirmed.
Thesis (Ph. D.)--University of Florida, 1987.
Bibliography: leaves 221-224.
General Note:
General Note:
Statement of Responsibility:
by Dennis Shuman.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
000943148 ( ALEPH )
16818207 ( OCLC )
AEQ4836 ( NOTIS )


This item has the following downloads:

Full Text







Copyright 1987


Dennis Shuman


I feel honored and extremely grateful *to have worked

with and learned from Dr. Keith White, my cobchair and

research advisor. Dr. White's brilliance was a continual

source of _awe and inspiration-to me, and nmmbined with his

Spract=ial__zesonxcefu ness, taught me -much about conducting

research. However, the education I received.from him far

exceeded academia, for Dr. White is one of the most caring,

loving, and self-sacrificing human beings TI-have encountered

in my life. His dedication to his studentsehis colleagues,

and his wife, as well as the countless timesehe has

compassionately bent over backwards to hilpeand support me,

I will never forget. et-

I would also like to express my appreciation to my

:h.-hair, -Dr.- Keith Doty, for his encouragemenn:-in the'pursuit

of my degree, and to my committee members, D. Keith Berg,

Dr. Don Childers, and Dr. Bill Dawson, for their guidance

and support during my many years of association with them.

Additionally, I want to thank Dr. Shelly Iseimberg for

helping me maintain a semblance of perspective and sanity

through it all.

I am deeply indebted to the administrative staffs in

-both the Departments of Electrical Engineering and


Psychology, with special thanks to Jim Fitzgerald, Cheryl

Phillips, Dotty Starry, and Doris Thames. I want to

acknowledge the help provided by all my fellow-research

assistants, especially John Krantz, Pascal programmer and

"fellow swayer". To my subjects, all those friends who

voluntarily submitted themselves to my torture machine in

order that I might leave the ranks of student serfdom,


Finally, I want to give thanks to my children Yossi and

baby Meira for giving me the inspiration and need-to finish

being a student, to my wife Renee for putting up with -my

perfectionism, and to my parents who instilled this

questionable trait in me so that one day they might kvell

"my son, the doctor".








1.1 Postural Overview . .. 1
1.2 Previous Postural Stability Studies 4
1.3 Methods Used in Postural Studies .. 16
1.3.1 Measurement Issues . .. 16
1.3.2 Previous Postural Measuring Systems 18
1.3.3 Previous Stimulus Feedback Systems 23
1.3.4 Position Sensor System Overview 25
1.4 Proposed Study . .. 27


2.1 Data Acquisition Subsystem .. 33
2.1.1 DAS Hardware . 35
2.1.2 DAS Geometric Analysis .. 49
2.1.3 DAS Software . 55
2.1.4 DAS Characterization .. 63
2.2 Stimulus Control Subsystem .. 68
2.2.1 SCS Hardware ... ...... 68
2.2.2 SCS Geometric Analysis .. 69
2.2.3 SCS Software . 73
2.2.4 SCS Characterization .. 89
2.3 Stimulus Projection Subsystem .. 90
2.3.1 SPS Construction . .. 91
2.3.2 SPS Calibration and Performance 96
2.4 Feedback Stimulus Characterization .. 98


3.1 Subjects .. . 104
3.2 Procedure . . 105
3.2.1 Experimental Environment .. 105
3.2.2 Subject Information and Instructions 110
3.2.3 Trial Parameters . .. 111
3.3 Data Analysis . 115

5 DISCUSSION . . 136

5.1 Position Sensor System . .. 136
5.1.1 Performance . 136
5.1.2 Applications . .. 140
5.2 Experiment . . 141
5.2.1 Informal Observations . 141
5.2.2 Spatial Correlation and
Postural Performances . 142
5.2.3 Altered Visual Feedback .... 146
5.2.4 Absolute and Relative
"Retinal Image Motion" .. 152
5.2.5 Summary . .. 159


A DAS SOFTWARE . . .. 161

A.1 CTRL.BAS Code . .. 161
A.2 CTRL.BAS Explanatory Notes .. 163
A.3 DAS Memory Map . .. 165
A.4 CTRL.MLP Code . . 166



C.1 Circuit Description . .. .173
C.2 Calibration . ... 178



D.1 Operating System Interrupt Correction .
D.2 DSTIM.BAS Code . .
D.3 DSTIM.BAS Explanatory Notes .
D.4 SCS Memory Map . .
D.5 SCS Machine Language Routines .
D.5.1 DYNHAND Code... . .
D.5.2 ERROR.MLP Code . .
D.5.3 TERM.MLP Code . .
D.5.4 VAR.MLP Code . .
D.5.5 EQRUN.MLP Code . .
D.5.6 D/A.MLP Code . .
D.6 Stimulus Jitter Simulation .



. 180
. 182
. 186
. 191
S. 195
. 195
. 196
S. 200
. 201
. 202
. 204

. 213



. 180




Position Sensor System
Component Sections .

2. Acoustic Source Driver and
Synchronization Circuit Components
with Operational Waveforms ..

3. Microphone/Detector Circuitry
with Operational Waveforms ..

4. PSS Geometry: Head Localization
and Stimulus Angle Determination.

5. DAS Software Flowchart:
Online Operation . .

6. SCS Software Flowchart:
Online Operation for
Stimulus Feedback Trials. .

7. Lateral Sway Gains with
Positive Stimulus Feedback Gains.

8. Lateral Sway Gains with
Negative Stimulus Feedback Gains .

9. Lateral Sway Gains Referenced to
Stationary Stimulus Condition. .

0. Anteroposterior Sway Gains with
Positive Stimulus Feedback Gains .

1. Anteroposterior Sway Gains with
Negative Stimulus Feedback Gains .

.2. Anteroposterior Sway Gains Referenced
to Stationary Stimulus Condition.

3. Rotational Movement Gains with
Positive Stimulus Feedback Gains.





S 45

S 52

S 59

S 79

S 120

. 121

S 122

S 123

S 124

S 125

. 126





14. Rotational Movement Gains with
Negative Stimulus Feedback Gains. ... .127

15. Rotational Movement Gains Referenced
to Stationary Stimulus Condition. ... .128

16. D/A Converter Circuit Implementation
with Operational Waveforms .. 176

17. DAS Rotational Angle Data Distribution
with Stationary Headpiece for
Generation of Jitter Simulation. 209



CMTS Calibrated Mechanical Transport System used in the
characterization of the PSS.

Cl The stimulus microcomputer, a component of the SCS.

C2 The position microcomputer, a component of the DAS.

D/A Digital to Analog conversion.

DAS Data Acquisition Subsystem of the PSS.

$ A prefix indicating a hexidecimal number.

DPS Data Processing Subsystem of the PSS.

DSB Data Storage Block, a dedicated- area of memory in C2.

mic Microphone or microphone/detector.

PSS Position Sensor System.

rad A unit of length defined as the radius of rotation of
the cylindrical screen (component of the SPS).

SCS Stimulus Control Subsystem of the PSS.

sp Speaker, click source, or acoustic source.

SPS Stimulus Projection Subsystem of the PSS.

8 Stimulus angle or stimulus projection angle.

VIA Versatile Interface Adapter, an I/O control device
(Synertek SY6522A).

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy




May, 1987

Chairman: Dr Keith L. Doty
Cochairman: Dr Keith D. White
Major Department: Electrical Engineering

Postural stability is a function of several weighted

sensory inputs to a central neural controller. A new

laboratory tool was developed to help quantify body sway

performances in general and to explore the contribution of

visual input as a feedback control signal in particular. By

utilizing body sway data collected online for the control of

a surrounding pattern's movement, the visual feedback loop

was altered to permit further characterization of postural


A parallel processing microcomputer system measured

body sway in any direction, and generated visual stimuli

based on mathematical analyses of the incoming sway data.

Position and orientation determinations were accomplished by

ascertaining the transit times of sequential sound fronts

from head attached acoustic sources to an array of fixed

receivers surrounding the subject. With two sources

employed, sampling rates of 25 Hz were achieved with

position and orientation accuracies better than 1 mm and

0.20 within a volume sufficiently large to accommodate the

extreme limits of body sway. The system managed the

horizontal position of a vertical grid shadowcast upon a

vertically oriented cylindrical screen encompassing the

subject's field of view. When governed by the sway data,

the grid position response delay introduced was less than 90

msec, and the position accuracy was better than 0.50 over

its full 320 range with unity feedback gain.

A study of 33 subjects employing stimulus movements

governed by body sway, with feedback gains ranging from -2

to +2, demonstrated differential influences on sway

frequency spectra by absolute and relative (to those

normally expected) retinal image motions. Only stimulus

movements that were both spatially and temporally correlated

with sway behaved as true feedback control. Sway spectra

thus observed were ordered directly with positive feedback

gains and inversely with negative feedback gains.

Additionally, those spectra exhibited large peaks and

valleys that displayed frequency consistency across feedback

gain magnitudes but reversed with feedback gain polarity.

The importance of regarding sway energy above 1 Hz was




1.1 Postural Overview

People are inherently unstable. While this statement

might prove to be quite controversial to a developmental

psychologist, it would not even begin to raise the eyebrows

of a sensory psychologist. The problem is familiar to

anyone who has watched a child learning how to stand. The

most commonly used model of an upright person is an inverted

pendulum. In static mechanistic terms, a body will fall if

its center of gravity is positioned outside its base of

support. Even a person with a size 11 shoe still has a lot

to contend with. Fortunately, we are endowed with muscles

which can exert forces on our skeletal frames to keep us

upright in a kind of dynamic quasi-stability. However,

these muscles need to be kept informed as to which way they

should pull, with how much force, and when. The requisite

information helps provide us with what is referred to as

spatial orientation, the sense of where one's body is and

how it is moving in relation to its environment. This is


supplied by our sensory inputs which function as our

interface with our surroundings.

Although any sensory inputs can provide positional

cues, it is well recognized that the visual, vestibular, and

proprioceptive senses play the dominant roles under most

circumstances. These sensory inputs are processed by the

nervous system which, in conjunction with assumptions based

on past experience, provides our spatial orientation and

decides what course of action to pursue. In situations

where all the inputs are interpreted as being in

concordance, the nervous system generally performs its

function superbly. The various inputs are not necessarily

redundant, but are rather in harmony, in that they provide

different facets of the overall situation. In addition,

their different resolution and frequency response

(bandwidth) properties make their relative value vary in

assorted situations, and the nervous system uses its own

program to weight these sensory inputs accordingly.' The

problem arises when the received information is

insufficient, misleading, novel, or conflicting. It is then

that disorientation, and possibly even physical

disturbances, can occur. The degree of disorientation is a

function of the weighting enacted by the nervous system, and

this program is apparently very individualistic and

adaptable. It is difficult to get a good objective measure

of disorientation, but one method that has been commonly

used is to measure postural stability or body sway. This


leads to another problem which is how to quantitatively

define and measure this parameter? This is one of the

issues addressed by this paper.

With the ability to measure spatial orientation, it

becomes possible to evaluate the role of, and gain an

operational understanding of, any one sensory input. In

many situations, visual perception plays the dominant role,

and sometimes with counterproductive results. Most people

have experienced the power of vision in establishing spatial

orientation while sitting in a stationary vehicle and

watching the vehicle next to them as it begins to move.

Often, the overwhelming sensation is that they are the ones

that are moving. There are many places where it is possible

to view movies projected on 3600 screens, and it is not

uncommon for people to lose their balance and/or experience

motion sickness while watching these panoramas filmed from

moving vehicles. In both of these situations the visual

cues dominate the other conflicting sensory inputs,-

compelling a false sense of self motion. In contrast,

travel sickness experienced inside a moving vehicle results

from the dominating, misleading visual cue of a still

surround in conflict with vestibular cues conveying the true

situation. This predicament is of special interest in space

flight where gravity, and its relationship to the

surrounding environment, can no longer be treated as



Vision performs a direct feedback role when conveying

information about body sway. A person overlooking a vista

will exhibit an decrease in postural stability, and thus a

greater tendency of falling, due to the decrease in retinal

image motion that occurs during sway while viewing a distant

object. Similarly, a new prescription for corrective lenses

can degrade stability due to the altered relationship

between head movement and retinal image motion. A greater

understanding of the role of visual feedback in balance,

studied by non-invasive means, holds the potential of

revealing significant contributory factors which induce

disorientation and thus promote falling. It is hoped that

this study will suggest possible remedies to reduce the

influence of these factors.

1.2 Previous Postural Stability Studies

The fact that visual information plays an important

role in the regulation of posture has been well appreciated.

A person standing with eyes open will exhibit spontaneous

postural sway which will increase in amplitude by up to 50%

when the eyes are closed. An example of this kind of test

is the classic Romberg test, which has been used as a

diagnostic aid based on the fact that postural stability

also largely depends upon vestibular and proprioceptive

information. Patients with diseases involving these other


systems show markedly increased closed eye body sway as

compared to normal subjects. However, the general

applicability of the Romberg test has come under question.

The dynamics of the influences on postural stability is

exceedingly complex, as can be appreciated by reviewing the

disconnected and sometimes contradictory data published to

date. Some of this divergent data is apparently the result

of the subtleties of these influences in particular

experimental designs, as well as the lack of uniformity in

what parameters are measured and the methods for doing so.

As examples, Brocklehurst et al. (1982) reported not finding

a correlation between visual deprivation and postural sway,

and Litvinenkova and Hlavacka (1971, cited by Gantchev et

al., 1972) found that in dark adaptation, closing the eyes

actually reduced body oscillations. Another seemingly

surprising contradiction was uncovered by Nashner and

Berthoz (1978). In examining the transient responses of

subjects to an unexpected movement of a platform upon which

they were standing (which were termed "early motor

responses"), they found that no difference in postural

performance (as measured by the body pitch and the EMG

activity of gastrocnemius or ankle extensor muscles) between

normal eyes opened and eyes closed conditions. Only in the

condition of a visual field stabilized with respect to the

subject's head was there a degradation in postural

performance. Thus false visual information did affect

response, but the lack of visual information did not degrade


it. Apparently, the weighting of the visual information was

somehow being altered.

The interaction between these three (visual,

proprioceptive and vestibular) systems in the control of

postural stability is quite complex and not well understood.

It should be noted that other systems can provide

stabilizing cues which can be significant, especially during

the lack of information from these three major sensory

systems. Examples are somesthetic (e.g., via a blind

person's cane) and auditory cues. Some experimenters have

had their subjects wear headphones transducing white noise

to eliminate extraneous auditory cues. The relative

weighting by the body of these sensory systems is very

individualistic and situationally dependent. Lestienne et

al. (1977), using a moving visual scene to induce postural

instability, found little or no effect with 20% of the

subjects. Most of these subjects then exhibited the more

typical instability when asked to perform mental arithmetic

while the stimulus was present, demonstrating not only

individual differences but also the subtlety of the

weighting factors. This is probably related to what Witkin

(1959) refers to as the two personality types. Field

dependent subjects sense their orientation with respect to

their (visual) surroundings, have lower IQs and are more

dependent on others. Field independent subjects sense their

orientation with respect to their internal "feel" in

relationship to gravity (vestibular), and individuals tend


to become more field independent as they age. This might

help explain the strong correlation between age and

increased postural sway (Era and Heikkinen, 1985). As seen

above, the typical sensory weighting pattern of the field

independent subject can sometimes be altered by a mental

distraction. Dorman et al. (1978) discusses how a

distraction as well an anxiety, inattention or diminished

eye fixation, can degrade postural stability, and is

probably related to the alteration of sensory weightings.

Vidal et al. (1982) notes that even the instructions given

to subjects can alter sensory weightings and postural

performance. If a subject is told that he can grab a fixed

metal bar to restore his equilibrium if necessary there

can be even complete suppression of the normal stabilizing

motor reflex in the lower limbs, even though the cues

transmitted from peripheral sensory systems are unaltered.

In general, the studies performed on postural stability

involved the stimulation of one or more of the sensory

systems while either leaving the others intact or altering

them (usually to diminish their effect). When the resulting

spatial information signalled by the various sensory

channels is in conflict rather than accord, sometimes a

single sensory channel determines the apparent spatial

orientation of the body. Gonshor and Jones (1980)

investigated vision reversal produced by wearing dove prisms

to produce horizontal inversion and reversed relative

lateral movement of the visual image. In spite of


appropriate vestibular and proprioceptive feedback, there

was a strong hierarchical predominance of vision resulting

in a violent loss of balance. In contrast to this, Nashner

(1982) found that unexpected functionally inappropriate

visual orientation cues do not always play a major role in

determining anteroposterior postural performance. He

proposed that the rapid reweighting of sensory inputs is

established by a fixed higher level process. He postulated

that the short term weighting of proprioceptive and visual

inputs is a function of the congruence of each of these

senses with the inertial-gravitational reference provided by

the vestibular system and that conflicting inputs from these

two senses are quickly suppressed in favor of those

congruent with the vestibular reference. Patients with

vestibular deficits (Nashner et al., 1982) and children

below the age of 7 1/2 years (Forssberg and Nashner, 1982)

do not have this higher level vestibular function to

reweight sensory inputs when they are conflicting, and so

experience greater instability than when just deprived of

these sensory inputs.

The idea of some pre-established strategy for rapid

resolution of conflict based on an internal reference system

is also found in the concept of "efference copy." According

to this theory, knowledge of intended movements is used to

construct an "internal model" of orientation which is used

to distinguish between shifts in body orientation relative

to the surround and changes in orientation of external


objects relative to the body. Nashner et al. (1982) were

concerned that a limitation of this concept in explaining

the observed phenomena is when reorientation of the body

occurs without prior knowledge, as when due to the

inherently unstable characteristics of the body in the

presence of external perturbations. Incorporation of a

memory-like process and less reliance upon the intentional

nature of movements mitigates those concerns to some extent

(Held, 1961).

A moving visual scene can induce a sensation of

self-motion called vection. Dichgans and Brandt (1978)

state that it is very likely that the afferent physiologic

mechanisms involved with vection are identical with those

stabilizing posture. When a subject views a moving visual

environment, a conflict is induced between the changing

visual input, which can be interpreted as body movement, and

vestibular and proprioceptive information which detect no

movement. Often, this conflict leads to disorientation and

motion sickness as well as increased postural sway.

Lestienne et al. (1977) attributed increased postural

oscillations during a constant velocity moving visual scene

to the visual-vestibular conflict reducing their "weight

ratio" in favor of proprioceptive input. This enhancement

or increased gain of the proprioceptive feedback loop would

cause overcompensation and thus oscillation. Another

example of the interactions between systems during conflict

was shown by Lackner and Teixeira (1977). They found that


although visual stimulation usually dominates vestibular

input, if a strong (continuously changing) vestibular and

proprioceptive input is generated by continuous and periodic

head movement, the effect of the visual stimulation was

lessened. This they likened to the domination of other

senses over vision when the retinal image is stable. They

concluded that covarying patterns of sensory change from a

number of receptor systems allow a more accurate assessment

of ongoing orientation because of less opportunity for

adaptation when the input is changing and because a changing

sensory input is "weighted" more heavily in determining

ongoing orientation. For convenience, this is termed the

"stimulus magnitude effect."

Teixeira and Lackner (1979) also found that as the

velocity of the visual stimulation was increased the greater

frequency of voluntary head movements needed to prevent

visual dominance of apparent orientation. However, the

issue could be raised that moving the head also increases

retinal image motion. The situation is complicated by the

fact that sometimes retinal image motion is accompanied by a

lowered weighting of vision as was found by White et al.

(1980) during saccadic eye movements. This is necessary to

prevent loss of balance during eye movement that is

initiated by the nervous system. However, White et al.

(1980) also found that retinal image motion caused by

external perturbation of the eyeball is not accompanied by a

lowered visual weighting, and postural stability is


disrupted. This implies that it is not eyeball movement

itself that is affecting weighting but perhaps eyeball motor

commands. Self initiated head movements also result in

retinal image motions that need to be suppressed for

maintenance of balance. Interestingly, Teixeira and Lackner

(1979) disproved the possibility that inhibition of the

visual stimulation effects was due to suppression by head

motor commands during voluntary movement by replicating

their previous results with passive (mechanically

controlled) head movements. Also the fact that either

rotary or tilting head movements produced similar results is

noted. In the Gonshor and Jones (1980) study with optical

reversal of vision, they found that there was progressive

recovery of postural control after donning the prisms in

walking tests but not in the standing tests. Although they

did not understand why, perhaps this was due to increased

proprioceptive and vestibular input (during walking) having

the effect of reducing the weighting of visual input.

Another example of this "stimulus magnitude effect" on

relative weighting is shown by Lestienne et al. (1977). In

their study of the influence of linear motion of a visual

scene on postural sway, they found that the sway amplitude

was proportional to area of the retina stimulated.

Another factor that influences the relative weighting

of the various stimuli is the current state of postural

stability. By comparing postural performances of subjects

on a stationary platform and one perturbed to induce body


sway, Soechting and Berthoz (1979) concluded that visual

stimulus motion produces a greater effect when tested in a

dynamic condition of postural change than when tested in

isolation in a static posture maintenance condition. This

might be likened to the difference between transient and

steady state performance, both necessary for full

characterization of a system.

The effect of altering the relationship between a

subject's movement and the resulting retinal image motion

(visual feedback) has already been alluded to. The

condition of a stabilized visual field, used in many studies

(e.g., Nashner and Berthoz, 1978 and Vidal et al., 1982),

involved a visual surround which tracked the anteroposterior

motion of a subject's head, thus attempting to eliminate the

subject's visual sense of motion. The stabilized visual

stimulus was always initially more effective in degrading

postural stability than an eyes closed condition. Vidal et

al. (1982) concluded that the eyes closed condition-could

not be considered visual deprivation but rather a tactic

that reduces visual weighting, and that the stabilized

condition results in an important disorganization of

postural control. Nashner and Berthoz (1978) also tried

moving the visual surround in the opposite direction of the

subject's head to enhance the visual input (any

anteroposterior movement of the subject appeared greater

than it really was), with a resulting reduction in body

sway. In a study (not true visual feedback) by Soechting


and Berthoz (1979), the subject was accelerated in a cart

while viewing a stimulus simulating one of three conditions

corresponding to viewing just the motion of the cart, the

stationary ground, or twice the cart motion. The angular

displacement of body was similar in the first two conditions

but was much greater in the latter condition. Their

interpretation was that the first two conditions simulated

normally encountered experiences but that the latter

condition did not. In the Gonshor and Jones (1980) study

with dove prisms, retinal image motion resulting from

lateral body sway had the same magnitude but was in the

opposite direction from normal. This extremely abnormal

situation gave a subject the impression of moving in a

direction opposite of the body's actual movement, resulting

in severe instability. The relationship between retinal

image motion and head motion can be thought of as a retinal

image feedback gain. Gantchev and Koitcheva (1981)

artificially simulated this feedback by having a subject's

anteroposterior sway controlling the vertical movement of a

horizontal line on a monitor viewed by the subject. They

found that increasing the gain (the movement of the

horizontal line) resulted in a decrease in body sway. In

another study using the same system, Gantchev et al. (1981)

found that reversing the direction of motion of the

horizontal line did not significantly effect body sway but

that introducing a delay in the visual feedback did. The

degradation in postural stability was greater with a


320 msec delay than with a 1200 msec delay (although the

delayed feedback performance was still an improvement over

the no feedback condition). Their conclusion was that the

degradation was a function of the phase relationship between

the delay and oscillatory characteristics of the body sway.

In the study of postural stability, another factor that

needs to be accounted for is plasticity and adaptation.

Leibowitz and Post (1980) describe their theory of the

"two-visual systems" concept with a focal mode for object

recognition and identification, and an "ambient" mode which

mediates spatial orientation, locomotion and posture. The

ambient mode is highly plastic and adaptable. For example,

Nashner and Berthoz (1978) found rapid habituation of the

influence of discongruent visual stimuli on EMG activity of

leg muscles (used in assessing stabilization performance of

the body) during repeated exposures. Their study also

showed how the motor system reweights the visual, vestibular

and proprioceptive inputs to postural control following

changes in the conditions of stimulation. They concluded

that visual inputs exert functionally different influences

upon two postural control responses, early and late. The

early response is immediately attenuated (as seen in EMG

activity) upon reception of any unexpected discongruent

visual input until the conflict is resolved. The late

response is compensatory, optimizing postural control with

the given sensory inputs. A more recent paper (Nashner et

al., 1982) states that the rapid reweighting of sensory


inputs utilizing a fixed internal model of the sensory

context, is not adaptive in the true sense of the term.

Adaptive changes in strategy which accompany modification of

the internal model occur over a much longer time interval

when the model fails to distinguish among self and external

object motions within a new sensory context. Such

adaptation was observed by Gonshor and Jones (1980) when

subjects donned and later removed reversing prisms. An

unexpected result they found was a marked and rapid

deterioration of non-visual performance after the subjects

began wearing the prisms, indicating that during adaptation

the non-visual postural stabilizing systems were undergoing


In summary, this review of what is known about the role

of visual cues in the stabilization of posture presents a

lot of fragmented bits and pieces that have yet to be put

together to show the whole picture. Some of these pieces do

not seem to fit, perhaps owing to the variety of

experimental methods, the sensitivity and individuality of

subjects' sensory weighting processes, and different

interpretations of data. It is hoped that the new apparatus

introduced here, and the research made possible by its

development, will help in the clarification of these issues.


1.3 Methods Used in Postural Studies

As technology has advanced, it has continually opened

new doors of inquiry into the operation of nature's most

sophisticated systems. In this modern age of computers, it

is now feasible to build instruments that can interact

online with biological systems, as well as collect and

analyze hugh amounts of data about them. Such technological

innovations presumably can extract new understanding of how

nature normally performs its functions, and well as provide

new means of assessing defective behavior.

1.3.1 Measurement Issues

There is not general agreement in the literature as to

what is the most appropriate gauge of postural stability.

If the body is assumed to behave as a rigid inverted

pendulum, then measuring the movement of any part will

define the movement of the whole. In actuality, the human

musculoskeletal system constitutes an active articulated

mechanism of extraordinary complexity with its 240 or so

degrees of freedom powered by approximately 750 individual

muscles. The correlation between the movement of any two

parts is a function of the plane of interest, the stance of

the subject, and the interface between the subject and the

ground. Inherent individual differences and the


nonlinearity of the system further complicate the issue.

Thus the validity of any simplifying assumption is context

dependent. The appropriateness of any parameter as a gauge

of postural stability is also a function of the testing

dynamics. As examples, postural performance can be the

transient response to or recovery from a sudden perturbing

stimulus, the steady state response to an ongoing

stimulation or condition, or even the transient response to

an ongoing stimulus condition when some other parameter is

suddenly changed.

Since the stability of the body is related to the

position of its center of gravity relative to its base of

support, monitoring of this dynamic mathematical point has

often been used. This is not an easy task, especially since

the location of the center of gravity moves within the body

as its parts move relative to each other. Measuring the

movement of parts of the body involved with its stabilizing

feedback mechanisms has also been employed. The head

contains the sensory transducers for the visual, auditory,

and vestibular feedback systems, and the rotation of the

ankle joint is considered a major source of proprioceptive

information. Once the postural measurements have been made,

the collected raw data still needs to be collapsed in order

for evaluations and conclusions to be extracted. There is

generally a lack of uniformity in the literature as to how

this analysis is accomplished.


1.3.2 Previous Postural Measuring Systems

As an example of a simple yet coarse mechanical measure

of postural stability, Gonshor and Jones (1980) measured the

time a subject was able to stand on a narrow rail. Thus

their test is related to movement of the body's center of

gravity relative to its interface with the ground. Another

relatively simple mechanical method, one which records head

position, was the use of a hat with a spring loaded pen

pointing upward to draw on a sheet of paper suspended over

the head of the swaying subject (K.D. White, personal

communication, 1985). This sway pattern on the paper was

subsequently scored to evaluate postural stability, a method

both tedious and of limited accuracy. Possible methods of

scoring are measurement of the maximum radius of the pattern

along various orientations, measurement of the total area

encompassed by the pattern, and measurement of the total

locus of sway as performed by Dornan et al. (1978).- They

used a computer and two potentiometric linear displacement

transducers connected by spring loaded strings to the sacrum

and the greater trochanter of a subject to graphically

produce and measure the total locus of sway of a part of the

body that roughly approximates its center of gravity.

Gantchev et al. (1981) were concerned with anteroposterior

sway and used the total locus of sway, or "total way" (sic),

only in that direction as the performance parameter. This

was accomplished with another variation of a mechanically


connected transducer as described by Gantchev et al. (1972).

A point approximating the center of gravity on a subject's

body was connected to a tensiometric lever whose

deformations were converted to an electrical signal.

However in similar study employing the same apparatus,

Gantchev and Koitcheva (1981) used the mean amplitude of

sway as the performance parameter. Nashner et al. (1982)

measured the anteroposterior sway of a subject's "center of

gravity" with a single potentiometric linear displacement

transducer connected to the hips with a lightweight rod.

The performance parameter was calculated by numerically

integrating the rectified sway signal (with steady state

offset removed) over the trial interval. Reason et al.

(1981) performed the sway signal integration by using a

voltage to frequency converter and a digital counter. Vidal

et al. (1982) were interested in the effect of a stabilized

visual stimulus on body sway in the sagittal plane. This

was accomplished by using the output of a single

potentiometric linear displacement transducer connected to

the subject's head by a lightweight rod to both control the

anteroposterior movement of a visual surround and also as a

sway measure. A cart upon which the subject was standing

was suddenly accelerated and the transient sway response was

recorded. The maximum angular pitch of the body was used as

a performance parameter, and this was calculated using the

potentiometric output signal in conjunction with the

assumption that the body acts as a rigid inverted pendulum.


Others such as Lestienne et al. (1977) connected an angular

displacement transducer to the subject's calf to measure

anteroposterior ankle rotation, and this was used directly

to determine the maximum angular body pitch. Similarly,

Soechting and Berthoz (1979) used an angular potentiometer

connected by a rod and crank mechanism to a helmet worn by

the subject to measure the anteroposterior body pitch angle.

Such connected devices introduce the major problem of

mechanical loading, which gives the subject positional cues,

as well as having limited resolution and range and of being

mechanically unidirectional in most cases.

A more sophisticated but relatively expensive

mechanical system is the stabilimeter or force plate as used

by Lestienne et al. (1977) and others. The subject stands

on a rigid plate mounted on strain gauges and as his center

of gravity shifts, the forces exerted on each strain gauge

change correspondingly. However, this device only yields

quantitative information about movement of the body's center

of pressure location (i.e. the position of the ground

reaction forces acting on the feet) which is only loosely

related to the body's center of gravity location

(Koozekanani et al., 1980). This in turn is not totally

representative of the body's position since the body

generally does not behave as the one link inverted pendulum

model as is sometimes assumed (Stockwell et al., 1981). In

addition, the mass and mechanical impedances of the

stabilimeter limit the upper end of its frequency response,


and it is inappropriate to use in situations where it is

desired to induce controlled acceleration of the body via

the feet as an input parameter (Soechting and Berthoz,

1979). In spite of its frequency limitations, Lestienne et

al. (1979) and White et al. (1980) chose to examine the

stabilimeter data collected in their studies using Fourier

transforms. In the former study the power spectral density

of anteroposterior sway was presented, and in the latter

study the lateral sway data was normalized with respect to

individual differences by presenting spectral gains (ratios

of power spectral densities of test and baseline

conditions). Also in the latter study (one of very few

concerned with lateral sway), the subject was instructed to

stand on one foot to decrease stability and enhance the

effect of visual stimuli.

A system which permits measurement of head, neck, hip

and knee movement was described by Booth and Stockwell

(1978). Pin lights attached to appropriate parts of the

body are detected by a videocamera whose signal is connected

to an interface. The interface assigns each light in the

videofield an X and Y coordinate value which is fed into a

computer for later analysis. Computer based video systems

have greater accuracy then mechanical systems, but can be

quite costly due to elaborate hardware and software

requirements. The fundamental resolution limits of video

systems are pixel size and number of pixels per videofield.

If the video field must encompass the entire range of


possible sway then, for commonly available NTSC standard

video cameras, approximately 256 x 256 pixels limit the

resolution per field. The smallest resolvable movement is

therefore about 1/256 of the field extent in the shortest

("video vertical") direction. In the system of measurement

described below (in section 2.1) the minimum resolvable

movement is more nearly 1/1000 or less of the range of

possible movement, a significant refinement. Two such video

systems have been presented as doctoral dissertations by

Cheng (1974) and Andres (1979).

Another source of data used in determining the effect

of stimuli on postural stability is EMG activity of leg

muscles. The muscles used in different studies were the

hamstrings, the quadriceps, the gastrocnemius, the anterior

tibialis, and the soleus. These data have mainly been used

in assessing transient performance after a sudden

perturbation or change of conditions. Some of the

performance parameters extracted from the data are -(a) time

to initial response onset, (b) time to peak response, and

(c) the energy of a response during some time interval.

Even the same parameters are generally evaluated by

different methods (e.g., different signal filtering schemes)

in the various studies, a complication in the interpretation

of these psychophysiological response measures.


1.3.3 Previous Stimulus Feedback Systems

Most previous studies employing a visual stimulus

controlled by movement of the subject were concerned with

one spatial dimension only. Nashner et al. (1982) had a

visual surround which rotated in the sagittal plane about an

axis connecting the ankle joints and controlled by the

anteroposterior sway orientation of the hips with respect to

the ankle joints (as measured by a potentiometric linear

displacement transducer). The precision of the visual

feedback was quite coarse because of the assumption that the

body acts as a one link inverted pendulum. Since there was

no compensation for head movements, the feedback gain

provided by the apparatus was not a constant. Vidal et al.

(1982) avoided this problem by having the anteroposterior

displacement of the head (as measured by a potentiometric

linear displacement transducer) control the mechanical

displacement of a visual surround along a track similarly

oriented. However, the timing characteristics of their

system were not reported. In a kind of visual pseudo

feedback system used by Soechting and Berthoz (1979), the

velocity (in a fore-aft direction) of a visual pattern

projected on a horizontal surface over the subject's head

(the sole visual input due to the use of blinders) was

determined by the transient velocity pattern of a cart

within which the subject was standing. Thus the movement of

the visual stimulus was tightly correlated to the movement


of the subject's feet but not the subject's eyes (because of

the head movements of the subject). The visual feedback

stimulus used by Gantchev and Koitcheva (1981) and Gantchev

et al. (1981) was a horizontal line displayed on a monitor

positioned 1 m from the subject. The vertical movement of

the line was controlled by the anteroposterior sway of the

subject (measured at a point approximating the center of

gravity of the subject's body). The gain, polarity, and

response delay of the vertical movement of the line were

independently variable, within limits. Again, actual

movements of the head and eyes were neglected in formulating

the visual feedback conditions.

In contrast to the above situations, in the Gonshor and

Jones (1980) study the movement of the visual feedback

stimulus was not limited to a single dimensional component

of the subject's movement. Now the stimulus correlation or

gain was limited to a single value with sign reversal

achieved optically and virtually instantaneously.

Unfortunately, the limited visual field forced by the use of

prisms plus the crude response measure used in this study

limit the generality of their findings. Dove prisms were

mounted in goggles so as to produce horizontal inversion of

the optical image, which reversed the direction of movement

of the retinal image during any head movement in the

transverse plane.


1.3.4 Position Sensor System Overview

The Position Sensor System (PSS) has been developed as

an economical yet high resolution tool for measuring body

sway in three dimensions, performing mathematical analyses

of sway data, and generating visual stimuli to ascertain

their contributions to postural stability. It is a vast

improvement over other systems to date used to measure

postural stability, and should expand the scope of research

possibilities as well as being a readily affordable clinical

screening tool. The PSS as currently implemented is a

multicomputer based system composed of four major

subsystems: a Data Acquisition Subsystem (DAS), a Stimulus

Control Subsystem (SCS), a Stimulus Projection Subsystem

(SPS), and a Data Processing Subsystem (DPS). The DAS and

the DPS together can measure and analyze postural stability

independent of the stimulus subsystems. Also the SCS can be

programmed to generate independently and to control-the

projection of visual stimuli by the SPS; else the SPS can

function separately given appropriate analog control


The microcomputer based DAS is the hub of the PSS. Its

function is to sample continuously the position and

orientation of a standing subject's body within a finite

volume that is large enough to accommodate the extreme

possible limits of postural sway in any direction. It also

functions as the "master" control center to run experiments


and coordinate the operations of the other sub-systems. A

unique feature of the DAS is that its operation is based on

acoustics. The principle involved is that the distance from

a point sound source to a point sound receiver can be

determined by measuring the propagation time of a

transmitted sound front. A source's position in three

dimensional space can generally be ascertained by employing

three noncollinearly-fixed receivers. The DAS samples the

distances from acoustic sources secured to a subject's body

to fixed microphones positioned within the subject's


The SCS generates analog waveforms that drive the SPS.

Based on a separate microcomputer, it can run independently

of the DAS, be synchronized by the DAS, and/or use the raw

position data transmitted by the DAS. In the latter case,

the SCS computes the subject's position and orientation

online, and uses some specified function of the processed

data to control the SPS.

The SPS incorporates a vertically oriented cylindrical

screen within which the subject stands, and a projector

which is positioned along the cylinder's axis of rotation

above the subject's head. The projector is a shadowcasting

device which projects a vertical grating onto the screen

that totally encompasses a properly positioned subject's

field of view. The grating moves horizontally across the

screen as dictated by the SPS input analog control signal.


The DPS stores and processes the raw data, and allows

for offline file manipulation to repair data errors. Some

of the statistical analyses performed offline on the

processed data are position and velocity means and standard

deviations, spatial variance and covariance. The Fast

Fourier Transform is used to examine and compare the

frequency spectra of body sway data collected under

different stimulus conditions. The DPS can provide

numerical and graphical outputs to either a monitor or a


The PSS is a powerful, versatile tool which will permit

the investigation of many factors involved in postural

stability. It has already been used, not only in the

present studies, but also in a research project involved

with postural stability'during saccadic eye movements

(Krantz, 1985), and is presently being used in related


1.4 Proposed Study

It has been demonstrated that visual cues have the

potential of being the dominant or heaviest weighted sensory

input, even when they are in conflict with other cues,

resulting in postural instability (e.g., Gonshor and Jones,

1980). Yet it has also been shown that when a visual cue is

in conflict with vestibular information as an internal


reference, the visual weighting may be reduced (Nashner,

1982). Nashner (1982) has shown that even when a visual cue

is destabilizing, its weighting is also a function of the

other inputs to the system. Possibly there are important

.situational differences between the two studies.

Gonshor and Jones (1980) studied direction reversal of

the lateral movements of retinal images while Nashner (1982)

studied a stimulus moving in an anteroposterior direction

with resulting collinear sways. It is not presently known

whether there are separate mechanisms for postural stability

in these two planes or what the interactions between them

might be. Certainly different skeletal muscles are

involved. Also lateral sway in front of an object results

in lateral retinal image motion while anteroposterior sway

results in a change in the size of the retinal image of the

surrounding objects (as well as a possible change in ocular


Another difference has to do with the visual feedback

gain, which can be expressed as the relationship between a

subject's movement and the resulting retinal image motion.

Normal vision provides negative feedback, defined as having

a retinal image feedback gain of minus one. In the Gonshor

and Jones (1980) study, the inverting prisms would have

produced a gain of plus one (direction reversal to positive

feedback) since movement of the subject resulted in an equal

but opposite retinal image motion from that expected. In

the Nashner (1982) study the stabilized (not environmentally


stationary) stimulus would have produced a gain of zero,

since subject movement results (in principle) in no change

in the retinal image. (Since there was no compensation for

head movements, the apparatus approximated poorly a gain of


Previous studies that attempted to vary systematically

the gain and phase of the retinal image feedback (Gantchev

and Koitcheva, 1981; Gantchev et al, 1981) have suffered

from contrived methods to control the (normally

uncorrelated) longitudinal retinal image motion by

anteroposterior body sway, and this precluded the

possibility of distinguishing between positive and negative

feedback. In addition, the feedback stimuli used only

subtended a small fraction of the total visual field (e.g.,

a video monitor positioned 1 m from the subject).

The present study proposes to quantify the effects of a

range of experimentally altered visual feedback gains on

postural stability, to clarify issues described above.

Postural performances will be expressed as comparative body

sway frequency spectra similar to those presented by White

et al. (1980). Frequency analyses of sway have seldom been

investigated, and when they have, they generally have

focused below 1 Hz since that is where most of the sway

component spectral energy resides. The common assumption

that nothing of interest occurs above 1 Hz will be explored.

The hypothesis is that visual feedback contributes to

at least two factors affecting postural stability; namely


the size of the retinal image motion and the correlation of

this motion with the body sway. The interaction of these

factors should be revealed by the influence of various

altered feedback gains on a subject's stability.

The first factor is related to how the weighting of

retinal image motion may be proportional to the magnitude of

the sensory input, so that altered visual feedback gain may

disrupt postural stability differently in the presence of

larger retinal image excursions ("stimulus magnitude


The second factor is related to how feedback affects

the stability of a system. It is further hypothesized that

the postural control system adaptively adjusts some internal

feedforward gain characteristic, so as to achieve

functionally appropriate body sway performance with the

inherent optically determined visual feedback gain. This

gain is proportional to the ratio of the magnitudes of

retinal image motion to head movement and is assumed to be

negative. Additional parameters that enter into the setting

of the internal feedforward gain are (a) the physical

characteristics of the body (e.g., mass distribution, muscle

tone, etc.), (b) the retinal, neural, and motor activity

time delays, (c) the environmental and situational factors

the body is normally exposed to (e.g., shoe type, gymnastic

training, mountain climbing, etc.), and (d) the internally

programmed relative weighting of visual and other sensory



Increasing the externally imposed negative feedback

gain might decrease body sway or some frequency components

of that sway, depending upon the phase lags introduced by

the experimental apparatus. Beyond a certain point,

increasing the negative feedback might lead to

overcompensating postural movements leading to increased

instability. Decreasing the negative feedback but

especially introducing positive feedback (direction

reversal) should be highly disruptive to the maintenance of


Another consideration involved with visual feedback is

the absolute and relative values of the input signal and how

each affects stability. If retinal image motion per se is

the sole visual feedback input to the system, then for a

given body movement, zero retinal image motion opens the

loop so that retinal image motion in the direction normally

encountered constitutes negative feedback and in the other

direction constitutes positive feedback. Therefore, for a

given body movement, relatively small retinal image motions

in these opposite directions may result in profoundly

different postural performances (notwithstanding reduced

visual weighting in both cases due to the stimulus magnitude


On the other hand, the system is "used to" a particular

value of negative feedback gain. If this value behaves like

a set or transition point, then a small change in gain from

above to below this value may result in a greater variation


in postural performance than would be found with a similar

change in feedback gain about a different quiescent point.


Overall design of the Position Sensor System (PSS) has

been described in the context of previous measuring devices

and of the proposed study in section 1.3.4. The complete

PSS consists of four subsystems, whose flow of control

generally presumes that the Data Acquisition Subsystem (DAS)

is the system "master" and that the other subsystems act as

its peripherals. The Data Processing Subsystem (DPS; a

TERAK 8510a and 8512 microcomputer with disk drives) simply

receives serial data transmissions for mass storage and

offline analyses. The Stimulus Control Subsystem (SCS) can

stand alone, but otherwise relies upon online data

transmitted via a parallel link to the DAS, in order to

control the Stimulus Projection Subsystem (SPS).

2.1 Data Acquisition Subsystem

The DAS is designed to ascertain the distances from

acoustic sources secured to a subject's body to fixed

microphones positioned around the subject. It is



predominantly based on hardware component sections that are

controlled by a small microcomputer which also functions as

the main control center for the overall PSS. Preliminary

versions of the DAS, presented in two previous papers

(Shuman, 1981a and Shuman 1981b), demonstrated that the

apparatus was capable of achieving a relative source to

receiver distance resolution of better than 1 mm over more

than a 1 m range.

In its present implementation the DAS is capable of

sampling distances at a 50 Hz rate. It generates pulses

which are gated to the acoustic sources in sequential order

to produce click sounds. Each of these pulses also starts

all the digital timers, one dedicated to each a microphone.

When a click sound is received by a microphone, its timer is

stopped, and after the click sound has been received by all

the microphones, the microcomputer reads and resets all the

timers. Acoustic source identification and synchronization

is achieved by click pulse width encoding and detection. If

operational errors are not detected, the DAS transmits the

collected raw data online to storage for offline processing,

and if appropriate, to the SCS for a stimulus position


2.1.1 DAS Hardware

The DAS hardware, represented in the PSS block diagram

in Figure 1, is composed of three main component sections

centered around the position microcomputer (see section

1.3.4 for a general system overview). In this depiction,

the sound generation component consists of an acoustic

source driver circuit and 1 to M sound transducers. The

sound reception component consists of 1 to N

microphone/detection circuits which are connected to the

computer via the timer interface circuit component. This

component contains 1 to N timers, a square wave generator,

test control and synchronization circuitry.

A subject could theoretically wear any number of

miniature acoustic sources, each determining the location of

that part of the body to which it is attached. In reality,

any obstacle in the path between an acoustic source and a

microphone, such as the subject's body, results in a

shadowing effect which alters (increases) their relative

distance reading. In addition, acoustic sources and

microphones have nonuniform directional characteristics

which become more problematic at higher frequencies. This

necessitates employing additional appropriately aimed fixed

microphones around the subject so that each acoustic source

has a clear path to a least three microphones for any

possible position of the subject. Also each source to

microphone path needs to be within some specified angle from




L L 'Z. --

---- ----- "--- ; ----' 1--- I-'- --1 L'" r 1 I U FULL BUS o\

Figure 1. Position Sensor System ComponentART Sections. These sections are the DRER




Figure i. Position Sensor System Component Sections. These sections are the Data
Acquisition Subsystem (DAS), the Stimulus Control Subsystem (SCS), the Stimulus
Projection Subsystem (SPS), and the Data Processing Subsystem (DPS).,


the central axis of the source, this angle being a function

of the source's dispersion characteristic and of the

frequency spectrum of the employed sound. It is then the

task of software, using error detection techniques, to

select the appropriate microphones to use in the

localization of each acoustic source.

For this study, the determination of the position and

orientation in the transverse plane of the head of a

standing human subject is used as the measure of postural

stability and in the derivation of the visual stimulus

control signal. This is accomplished by the use of two

sources attached to the head and positioned along a

transverse plane. In order to achieve adequate

orientational resolution and to be outside of the subject's

field of vision, the sources are positioned 40 cm apart on

opposite sides of the head. This is implemented by

fastening them to the ends of a slat (constituting the

"headpiece") which is attached to a hat consisting of the

plastic webbing taken from the inside of a construction hard

hat. The webbing can be adjusted to fit securely to

eliminate relative movement between the sources and the

subject's head. Four microphones are positioned at the

vertices of a square of length 100 cm mounted horizontally

7'-8" above the floor. This square, bolted to the ceiling

and referred to as the system overhead frame, is constructed

from aluminum L stock with flat aluminum diagonal

crosspieces to maintain squareness and support SPS projector


and screen. To collect data, a subject is initially

positioned under the center of the square and facing

parallel to two of its sides. The acoustic sources point

upward and the microphones are aimed inward towards the

center of the square and downward at a 45 degree angle to

the plane of the square. This configuration works within

the limitations of the directional characteristics of the

acoustic sources and the microphones for any expected body

sway of the subject.

The volume of space within which acoustic sources could

theoretically be localized is anywhere between the plane of

the suspended square and the floor. With the practical

constraints of the amplitude of an acoustic source output,

the sensitivity of the microphones, and noise

considerations, it was necessary to define a volume of space

within which the acoustic sources could be expected to be

found. A decision was made to impose the limitation that

the distance from the top of a subject's head (where the

acoustic sources are positioned) to the floor be within the

range of 5' to 6'-4". It was empirically determined that

the maximum a subject's head could move in any direction

along the transverse plane during body sway was 30 cm. This

information was used in calibration and characterization of

the DAS as will later be seen.

The acoustic sources are common miniature earplug

speakers modified to broaden their sound dispersion

characteristic by cutting off the protruding plastic that


would normally fit in the ear. Each source employs an eight

ohm voice coil to move a small ferrous diaphragm and the

sound is emitted through the 7 mm diameter hole left after

modifying its plastic case. Under a worst case condition of

a subject's head being 6'-4" above the floor and swaying

30 cm away from a microphone, the maximum angle between the

central axis of the distal source and a line connecting this

source/microphone pair is 70 degrees. Using data previously

presented (Shuman, 1981b), this results in a maximum

increase of 3 mm in the sampled distance beyond the value

found when the source is pointed directly towards the

microphone. The relative error between consecutive samples

for any possible body sway is substantially less.

The sources are powered by the acoustic source driver

circuit represented in Figure 2. The complete schematic

diagram for the present implementation was previously shown

in Shuman, 1981b. The driver generates pulses which are

gated sequentially to the appropriate sources to produce

click sounds. The circuit also generates a timer control

signal (summed click pulses) to notify the timer circuits of

each click sound's transmission. In the initial design

(Shuman, 1981a), the battery powered driver circuit was to

be worn by the subject and the timer control signal would be

relayed by an FM transmitter/receiver pair to the timer

circuits. This was to insure that proprioceptive positional

clues would not result from wires connecting the subject to

any stationary object. In the present implementation, a

60 80

sec sec

7O/1 70/1 1
()- seq sec
I 0 1



Figure 2. Acoustic Source Driver and Synchronization
Waveforms. Acoustic source identities are encoded
synchronization. The acoustic source driver pulse

Circuit Components with Operational
in the click pulse widths for
has a slow decay to reduce ringing.



stationary driver circuit is connected to the acoustic

sources by fine wires dangling from the end of a light

spring over the head of the subject. This is found to also

adequately eliminate proprioceptive clues and obviates the

use of the FM transmitter/receiver pair. For safety

considerations, the driver circuit is still battery powered

so that the acoustic sources attached to the subject are

never even indirectly connected to the 120 volt power line

in the laboratory. The batteries in the driver circuit (two

9 volt transistor batteries) are automatically recharged

whenever the main power to the PSS is turned off.

The acoustic source driver circuit can operate in three

possible modes: a free running sequential order mode for

normal use and two manual modes for testing and calibration

(a sequential order mode and a single acoustic source

repetition mode). The free running mode requires a square

wave input signal which can be generated by a self contained

oscillator whose frequency is controlled by a precision

variable resistor. In the present implementation, the

square wave input is more precisely generated and controlled

by the timer interface circuit under software control. The

acoustic source pulses are at the oscillator frequency and

so the DAS sampling frequency is equal to the oscillator

frequency divided by the number of acoustic sources. The

manual modes use a mechanical switch and a debouncing

circuit to generate single pulses.


Each pulse of the square wave input signal train is

sequentially gated to the appropriate power circuit

dedicated to a particular acoustic source. A power circuit

consists of a EGC 186A power transistor whose base is driven

by a dedicated one shot. In order to create a sharp click

sound (i.e., a fast rise time sound wavefront), the one shot

output biases the power transistor into saturation which in

turn supplies the acoustic source with a fast rise time high

current pulse (approximately 2 amps). The diode across the

inductive voice coil of the speaker reduces the voltage

induced when the transistor is turned off. It was

empirically determined that a one shot pulse width between

approximately 50 to 80 usecs results in a maximum amplitude

microphone output for the transducers used. This is

apparently due to the resonant frequency of the microphone

which rings at about 8 kHz since half of its resonant period

is approximately equal to the mean of the above one shot

pulse width range. Such a short pulse also does not

overheat or burn out the acoustic source for the duty cycles

involved (less than 0.25% with a 25 Hz sampling frequency).

It was discovered that the rectangular current pulse

resulted in the click sound persisting for a long time after

the pulse ended, due in part to the resonant oscillations of

the speaker's diaphragm. This in turn reduced the maximum

sampling frequency of the DAS. The problem was partially

solved by slowly returning the speaker's diaphragm to its

rest position. This was accomplished by reshaping the


rectangular base drive of the power transistor to be more

like a sawtooth. An RC circuit providing a 2 msec decay

time constant was empirically found to permit the highest

click rate frequency.

The outputs of all the one shots are summed to provide

the timer control signal previously mentioned. In addition,

this composite signal is used by the position microcomputer

to achieve synchronization with the acoustic sources by

means of the sync detector section of the timer interface

circuit. As shown in Figure 2, the one shot output is about

60 Isec for acoustic source 1 and about 80 usec for all the

other acoustic sources. The composite signal, shown as

waveform A, triggers the sync detector one shot which has an

output duration of about 70 usec, shown as waveform B. The

sync detector latch, which is reset by the position

microcomputer at the leading edge of each acoustic source

one shot output, is set by a latch pulse waveformm C) which

is only generated when acoustic source 1 is activated. The

position microcomputer reads the latch to ascertain the

beginning of a sequence of acoustic clicks and also to check

that the number of clicks in each sequence (corresponding to

a complete sample) is correct, as is discussed further in

section 2.1.3 on the DAS software.

The function of each of the microphone/detection

circuits is to indicate precisely the time of reception of

an acoustic click transmitted from anywhere within the

volume of space that the subject's head could reasonably be


located. Each circuit, as portrayed in Figure 3, is self

contained in a shielded metal-box. The microphone sound

transducer is a miniature piezoelectric crystal element

which produces a ringing response at its resonant frequency

(approximately 8 kHz) when the click sound pressure wave

front impinges on it. This ringing is first amplified by a

differential input 40 dB preamplifier utilizing an LM725

operational amplifier with external frequency compensation.

This op amp was chosen because of its wide bandwidth and

good noise performance as well as its high common mode

rejection ratio (CMRR) of 120 dB. The common mode rejection

(CMR) of the circuit is maximized by adjusting RCOM. A

second stage of about 50 dB gain employs an externally

compensated LM725 op amp in an inverting amplifier

configuration to bring the signal level up for detection.

The two stages are directly coupled and their net offset

voltage, as measured at the output of the second stage, is

nulled out by adjusting ROFF in order to maximize the

dynamic range of the pair. The signal is then capacitively

coupled (to remove any dc component due to drift) to a

threshold detector consisting of an LM741 op amp in an open

loop configuration with a threshold level set by RTH. The

output of the detector is a pulse which goes from about

-10.4 V to +10.8 V whenever the signal crosses the threshold

level. A common emitter transistor converts the pulse to

appropriate logic levels so as to act as a stop pulse for

the appropriate timer. Because of the finite slew rate of

Radio Shack 273-06

(Threshold Level Adj.)





OF "CLICK" 2nd



Figure 3. Microphone/Detector Circuitry with Operational Waveforms. The threshold level
is set to detect the third peak of the microphones ringing response to the click sound
produced by an acoustic source positioned within the operational volume of the DAS.


) ))



the detector op amp, its output takes almost 30 usec to

swing from its negative to positive saturation values. This

characteristic is capitalized upon in the performance of

some low pass filtering which reduces false detection by

noise spikes. A 6.2 V zener diode is incorporated into the

base circuit of the transistor so that true detection

requires the op amp output to exceed about +6.9 V or more

than 75% of its possible voltage excursion. Therefore,

detection requires the input signal to exceed the threshold

level for almost 23 usec.

The initial ringing response to a click sound wave

front at the input of the threshold detector, as shown in

Figure 3, is a growing sinusoid whose peak amplitudes are a

function of the position of an acoustic source relative to

the microphone element. The threshold level should be set

high enough to eliminate false triggering due to electrical

and unavoidable audible noise and low enough to detect the

click sound when the acoustic source is at its expected

worst case distance and orientation relative to the

microphone element. Another consideration, previously

alluded to during the discussion of the acoustic sources'

driver circuits, is the effect of the threshold level on the

maximum sampling frequency of the DAS. The persistence of

the ringing response, which is a function of the click sound

waveform as well as the microphone element characteristic,

prohibits the transmission of the subsequent click sound

until the ringing has decayed below the threshold level.


Once below this level, a sizable ringing amplitude might

still alter the detection time of the contiguous click sound

as a result of constructive or destructive interference.

This is one source of error which affects the accuracy of

the DAS (discussed in section 2.1.4). An empirical study

was performed to determine which peak of the growing

sinusoid should be detected for an optimal performance

range. Earlier peaks provide a truer indication of the

click sound wave front arrival and fluctuations in the

ringing waveform introduce less error, but the larger

amplitudes of later peaks allow for a higher threshold

level. However, beyond the second peak, false detection by

a previous peak can easily occur when an acoustic source is

close to a microphone. A figure of merit for these later

peaks is the amplitude ratio of a peak with the preceding

peak of same polarity, and this was greatest for the third

peak. Within the volume of space of interest, the third

peak minimum amplitude was 0.88 V and the first peak maximum

amplitude was 0.27 V. Based on the discussed tradeoffs, a

threshold level of 0.7 V was selected. The polarities of

the acoustic sources (actually connected to the driver

circuits via polarity reversing switches since the

polarities of the miniature earplug connectors were found to

be random), microphone elements and threshold levels were

chosen to detect the third peak.

The timer interface circuit is entirely digital and

could be implemented in a variety of ways depending upon


which microcomputer is chosen for the DAS to be centered

around. The present implementation, discussed in detail in

Shuman, 1981b, uses a modified Ohio Scientific Challenger IP

8K microcomputer based on a 6502 processor running at 2 MHz.

The timer interface circuit is composed of a Synertek

SY6522A Versatile Interface Adapter (VIA) dedicated to each

microphone/detection circuit, supporting TTL chips, and

buffers for the computer's address bus. The VIA's "Timer 2"

16 bit timers count pulses from gated 1 Mhz signals derived

from the microcomputer's internal crystal clock giving a

detection digitizing resolution of 1 usec. These signal

gates are opened synchronously by the summed click pulses

output of the acoustic source driver and are closed by the

first output pulse of its dedicated microphone/detection

circuit. Using hardware timing permits greater accuracy in

measuring these asynchronous events. Software controls the

non time-critical functions of initializing and reading of

the timers. Modifications to the timer interface circuit

since the description presented by Shuman (1981b) are the

inclusion of test control, the square wave signal generator,

and the parallel link. Test control consists of begin and

halt mechanical grounding switches (pulled up for noise

protection) interfaced to the computer by port lines PB4 and

PB3 on VIAl (the VIA dedicated to microphone/detector 1 or

"mic 1") and monitored by software. The square wave

generator utilizes "Timer 1" on VIAl in its free-run mode

which is set to frequency, started, and stopped under


software control. The output of this generator timer, on

PB7, is stepped down to the desired range of frequencies for

the acoustic source driver by a TTL 7493 divide by 16 chip.

The parallel link, which transmits the microphone dedicated

timers' readings to the stimulus microcomputer, utilizes a

byte wide port (PAO-PA7) on VIA2 for data and its two

associated control lines (CAl and CA2) for handshaking. The

port is connected, with a one to one pin correspondence, to

an equivalent VIA port in the SCS. The control lines are

cross connected to the VIA control lines from the SCS (input

control line CAl on VIA2 to output control line CA2 from the

SCS VIA and vice versa). The sync detector, discussed

previously, is also part of the timer interface circuit with

its latch reset and read by VIA1 under software control.

2.1.2 DAS Geometric Analysis

The following geometric analysis, although not directly

performed or used by the DAS, is presented at this time to

clarify how the DAS works in conjunction with the rest of

the PSS to fulfill its role in measuring postural stability.

The mathematical results of this discussion are actually

utilized in the data analysis performed by the SCS and the


If the distance from a movable point sound source to a

fixed point sound receiver is ascertained, then the source


is known to be located somewhere on the surface of a sphere

centered about the receiver with a radius equal to that

distance. If the distance to a second receiver is known,

then the source's position is somewhere along the

intersection of the two spheres, which is a circle whose

axis of rotation is the line connecting the two receivers.

If the distance to a third receiver not on line with the

other two is known, then the source is at an intersection of

this new sphere with the previous circle, which is two

points equidistant from and on opposite sides of the plane

defined by the three receivers. If it is known which side

of the plane the source is on, then the location of the

source in three dimensional space relative to the receivers

is established.

For this study, the determination of the position and

orientation of a subject's head along a transverse plane,

which involves two translational and one rotational degrees

of freedom, is deemed sufficient (Krantz, 1985). This can

be ascertained by minimally employing two acoustic sources

attached to the head along a transverse plane. An

unrestrained head can move with all three translational and

three rotational degrees of freedom. With the two sources

fastened on opposite sides of the head, all these degrees of

freedom can be resolved with the exception of rotation about

a line connecting the two sources (like nodding "yes").

However, during typical sway, a body does not lean beyond an

amount that would displace its center of gravity beyond the


support base of its feet. Therefore, the up and down motion

of the head can be considered a second order effect, and

rotations within the sagittal and coronal planes yield

little additional information beyond that obtained from head

translations in the transverse plane. In addition, within

the context of this study employing a vertical grid stimulus

constrained to moving laterally around a cylindrical screen,

information about head movement involving these other

degrees of freedom could not, in any event, have any effect

on the stimulus movement.

Figure 4 presents a graphical depiction of the

geometric relationships between the two acoustic sources and

four microphone/detectors projected on to a transverse

plane. The subject is standing at the intersection of lines

L1 and L2 looking at a point (Yp,Xp) on the cylindrical

screen. The two acoustic sources, Sp 1 and Sp 2, are

positioned on opposite sides of the head at the ends of line

LI. The four microphone/detectors, Mic 1 through Mic 4, are

positioned at the vertices of a square that is over the head

of the subject, thus removing any ambiguity as to which side

of the receiver plane the source is on (as discussed above

in this section). The distances, D1 through D8, correspond

to the eight possible speaker-microphone (sp-mic) pair

combinations. These distances are used to calculate the X

and Y coordinates of Sp 1 and Sp 2 using the following

equation derived from the Pythagorean theorem in Krantz


Figure 4. PSS Geometry: Head Localization and Stimulus
Angle Determination. The subject's head is located at
the intersection of line LI (connecting the acoustic
sources) and line L2 (a segment of which is the subject's
line of sight). The distances from acoustic sources to
microphones (Dl through D8) are used to calculate the
acoustic sources' coordinates which defines the subject's
location and orientation. The intersection of the
subject's line of sight with the cylindrical screen is
used to calculate the stimulus visual feedback angle e.


(1) CV = ((Dist A)2-(Dist B)2)/(2L)

where CV is a coordinate value of an acoustic source, Dist A

and Dist B are the distances from this acoustic source to

two microphone/detectors lying on a line parallel to the

axis that is the same as the desired coordinate, and L is

the distance between those two microphone/detectors

(empirically found to be 100.2 cm). Thus, in conformity

with the above discussion, the determination of the location

of an acoustic source in a transverse plane requires the

employment of three microphone/detectors. With the four

available microphone/detectors, each of the four coordinate

values can be calculated using two different distance pairs,

and this redundancy allows for a certain degree of error

correction. The choice of which distance pair to try first

is based on the practical consideration of minimizing

potential interference. For the X coordinates, the

determining factor is possible head shadowing, and for the Y

coordinates, it is possible ambient noise coming frbm behind

the subject (due to the location and orientation of the

experimental setup relative to the experimenter and the rest

of the laboratory). Table 1 presents the resulting default

and alternative coordinate determinants. For example, the

default implementation of equation (1) for the Y coordinate

of Sp 1 is

(2) Y1 = ((D2)2 (D1)2)/(2L)

and the alternative implementation is

(3) Y1 = ((D3)2 (D4)2)/(2L)


Table 1
Coordinate Determinants

Speaker Default Alternative
Coordinate Distances Distances
Y1 D2-D1 D3-D4

Xl D4-D1 D3-D2

Y2 D7-D8 D6-D5

X2 D8-D5 D7-D6

A raw position datum for a sp-mic pair is a 16 bit

integer corresponding to the number of (approximately) 1 Mhz

clock cycles that occurred between the beginning of the

acoustic source driver pulse and the detection of the third

peak of the microphone/detector's ringing response (see

section 2.1.1). The expression of the relationship between

this number and the distance from the acoustic source to the

microphone detector is

(4) C = mD + b

where C is the number of clock cycles, D is the distance, m

is a factor related to the speed of sound in air, and b is a

constant resulting from the delays involved with the

transmission and the detection of the sound front. It was

empirically determined that for all the sp-mic pairs

m = 28.46 cycles/mm with a variance of 0.0006

and b = 175.5 cycles with a variance of 31.0

which is reasonable since m is a physical constant and b is

a function of variability in electronic components.


Combining the above equations and empirical measures, the

expression for a coordinate value in mm is

(5) CV = ((CA)2 (CB)2 + 351(CB CA))/16232

where CA and Cg are the sp-mic pair raw position data

measured in clock cycles, using the pair distances described

in Table 1.

2.1.3 DAS Software

The DAS software is an online system that manages the

execution of experimental trials, which in turn involves the

collection and transmission of raw position data as it is

generated. The raw datum for each sp-mic pair is a 2 byte

number read from the timer dedicated to that particular

microphone/detector. In this implementation, a complete

sample is composed of two groups (corresponding to the two

acoustic sources) of 8 bytes (corresponding to the four

microphone/detectors) for -a total of 16 bytes. DAS

management involves control of the sampling rate, number of

samples per trial, and initiation of a trial. In addition,

the program checks and corrects for loss of acoustic source

number synchronization and overflow conditions which occur

if the input data rate exceeds the output rate. It is

designed to run in conjunction with the previously described

DAS hardware (section 2.1.1) centered around a 6502 based

modified Ohio Scientific Challenger CIP microcomputer

(henceforth referred to as C2). In order to perform its

function at maximum speed, the program runs at a machine

language level uhder the control of a BASIC program which

also performs offline tasks and acts as a user interface.

The complete program fits well within the 8K of user RAM,

with less than 350 bytes allocated to the machine language

program and associated buffer locations.

There are presently several different versions of this

software in _each of two groups. The various BASIC versions

provide for different experimental designs and for either a

software or hardware controlled square wave generator input

to the acoustic source driver. The two groups have to do

with the machine language portions of the programs. Group 1

is optimized for the highest possible sample rate, with

transmission of the raw data via an RS-232 4800 baud serial

link to a mass storage device (the DPS) for offline

analysis. This is accomplished by transmission of the

collected raw data from the previous sample during the

software idle periods between transmission and reception of

the click sounds (typically 2 to 4 msec). Group 1 programs

can Operate at a 50 Hz click rate or a 25 Hz sample rate

with a two acoustic source implementation. Group 2

programs, one of which was used in this study, are designed

for the case in which the motion of the subject can affect

the stimulus motion. Therefore the raw data are transmitted

to the stimulus microcomputer (henceforth referred to as Cl)

via a byte wide parallel data link as well as to the DPS.


In this case, the delay until transmission of the raw data

is crucial and so this delay is minimized at the expense of

reducing the maximum click rate to 20 Hz or a 10 Hz sample

rate. In order to avoid having the position microcomputer

(C2) wait for the stimulus microcomputer's (Cl) need for

more data, any attempted transmission by C2 on the parallel

link generates an interrupt on Cl.

The BASIC program used in this study, CTRL.BAS, and the

assembled form of its group 2 machine language program,

CTRL.MLP, along with detailed documentation and a memory map

of the system, are all included in section A in the

Appendix. Operation of the DAS software begins with

CTRL.BAS poking (loading) the machine language program into

memory, and then initializing the timer interface circuit

and locations in C2's RAM involved with the operation of

CTRL.MLP. Of major interest are the sync memory location

which keeps track of which acoustic source click sound is

next expected, the sample counter which monitors the

duration of a trial, and the data storage block (DSB) which

is an 8 byte buffer used for temporary storage of each data

group before transmission via the serial and parallel data

links. The timer interface circuit initialization includes

the (a) set up the parallel link to interrupt Cl when C2 is

ready to transmit data from the DSB, (b) notify Cl when

valid data is on the link, and (c) wait until Cl

acknowledges receiving each byte before sending the next

one. If acknowledgement is not received within certain time


period (the time until the next click sound is transmitted),

a flag is set and transmission via the parallel link is

aborted. The raw timer data are stored in the DSB prior to

transmission in order to have a briefer interrupt of C2 than

would be possible if the individual timers were read and

written directly to the parallel link.

After initialization is completed, the user is asked to

enter the desired sampling rate and number of samples per

trial. A prompt on the monitor asks the user to turn on the

battery powered acoustic source driver, to check that the

stimulus computer is ready, and to press the BEGIN switch.

Pressing the switch turns on the square wave generator which

starts the click sounds and CTRL.MLP is entered. An

overview of CTRL.MLP is shown in the flowchart in Figure 5.

The system is prepared to expect a click sound from Sp 1 by

setting the sync memory location and the countdown timers

are initialized (to $FFFF, where $ indicates an hexidecimal

number) in preparation for counting clock pulses. The HALT

switch used to abort manually and the sample counter which

terminates a trial if it reaches the user specified number

of samples per trial are checked. If termination is

indicated, a termination routine is entered which begins by

loading the DSB with a termination code (8 bytes of $FF). A

delay insuring that Cl is ready to receive data (Cl

operation is discussed in section 2.2.3) is accomplished by

waiting for the next click to be transmitted and detected by

Mic 1. The termination code in the DSB is transmitted a

The Data Storage Block (DSB) C Enter from BASIl
accumulates all the timer data
for each click prior to transmission
to the Stimulus Computer (C1)
and Mass Storage. I nitialize for Sp

Figure 5. DAS Software Flowchart: Online Operation.
CTRL.MLP manages the detector timers, acoustic source
synchronization, and timer data transmission.


byte at a time to Cl via the parallel link, then serially to

mass storage, and then again to Cl. Cl may need to receive

the termination code twice before recognizing termination by

C2 (depending upon the state of Cl at the time of

interrupt), and the two interrupts of Cl need to be

separated by a time period provided by the transmission to

the DPS (discussed further in section 2.2.3). The sample

counter is updated, then the program returns to CTRL.BAS

which turns off the square wave generator, thus stopping the

clicks sounds, and displays the number of samples in the

trial on the monitor. After a pause, CTRL.BAS is ready for

the next trial and again prompts the user to turn on the

acoustic source driver and check that Cl is ready. The

sampling rate and number of samples per trial can only be

specified before entering the first trial. If the BEGIN

switch is pressed before the acoustic source driver is

turned on, a message to that effect is displayed on the

monitor and, after a pause, CTRL.BAS is again ready for the

next trial.

If a termination (manual abort or sample count reached)

was not indicated during the running of CTRL.MLP, it then

waits for a click sound (from Sp 1 if CTRL.MLP was just

entered) to be transmitted. Upon transmission, which starts

the timers counting (purely a hardware function), CTRL.MLP

clears the sync detector latch and waits for Mic 1 to detect

the click sound. Upon detection, which stops the Mic 1

timer (again a hardware function), CTRL.MLP performs a sync


check by comparing the expected acoustic source identity

(stored in the sync memory location) with the actual

acoustic source identity (as read from the sync detector

latch). If the sync check is good, the 2 byte datum from

the Mic 1 timer is stored in the DSB and the program

checks/waits for Mic 2 to detect the same click sound (which

it may have already). When it does, the datum from the Mic

2 timer is stored in the DSB and the same procedure is

followed for Mics 3 and 4. The sync check is only made one

time for each click sound, and that is after its detection

by Mic 1. Once all the raw timer data are stored in the

DSB, its contents are transmitted to Cl via the parallel

link. If data transmission is not successful, as evaluated

by the handshaking, the termination routine is entered. If

data transmission is successful, the DSB is then also

transmitted to the DPS via the serial link, the sample

counter is updated, and the sync memory location is set for

the next acoustic source. At this point, a check is made to

see if the next click sound has already been transmitted.

If it has not, CTRL.MLP loops back to prepare the timers for

the next click sound and the program continues. If the

click has been transmitted then CTRL.MLP has fallen behind

(an overflow condition), and the program returns to CTRL.BAS

which displays a message to that effect on the monitor.

After a pause, CTRL.BAS is again ready for the next trial.

Such overflow can be caused by too fast a sampling rate, by

missed detection of a click sound by one of the microphones,


or by excessive acoustic source to microphone/detector


If sync check fails then CTRL.MLP increments a sync

error counter displayed on the monitor and loads the DSB

with the sync error code (8 bytes of $00). If a click sound

from Sp 1 was expected but was received from Sp 2, then the

sync error code is sent to the DPS twice and the sample

counter is updated. This fills a complete record in mass

storage so that it has an indication of the sync error and

is also prepared next to receive timer data corresponding to

Sp 1. Nothing is sent to Cl so that the stimulus position

is not updated and Cl is still expecting timer data

corresponding to Sp 1. The program then continues by

waiting for a click sound to be transmitted, with the

expectation that it will be from Sp 1. If the sync error

occurred when a click sound from Sp 2 was expected but was

received from Sp 1, then the sync error code is sent to Cl.

This aborts the next stimulus update even though va-lid timer

data corresponding to Sp 1 is currently held by Cl (see

section 2.2.3), and leaves Cl expecting timer data

corresponding to Sp 1. Under this sync error status, if

data transmission to Cl is unsuccessful, then the

termination routine is entered. If data transmission is

successful, then the sync error code is then sent to the

DPS. This completes the partially filled record in mass

storage so that it has an indication of the sync error and

is also prepared to next receive timer data corresponding to


Sp 1. The sample counter is then updated. Now Cl and mass

storage are expecting timer data corresponding to Sp 1 so

that the next click sound needs to be ignored. After all of

the mics have detected this next click sound and all timers

stop, the program loops back to set the sync memory location

for Sp 1 and continues as previously described.

2.1.4 DAS Characterization

To fully and accurately characterize the DAS (and later

the SCS), it was necessary to be able to translate and

rotate the headpiece (the two acoustic sources attached to

the supporting slat) over the DAS operational range with an

accuracy substantially greater than that of the DAS. In

order to accomplish this task, the Calibrated Mechanical

Transport System (CMTS) was designed and assembled

(described in section B in the Appendix).

The DAS, whose actual output is timer counts

corresponding to distances between microphone/detectors and

the two acoustic sources, was characterized based on the

processed data output of the offline DPS resulting from

discrete tests. For each trial, the DPS calculated the

means and standard deviations of (a) the Y and X coordinates

of the midpoint of a line (L1 in Figure 4) connecting the

acoustic sources, corresponding to the location of the

center of the headpiece, and of (b) the rotational


coordinate of a line (L2 in Figure 4) normal to line Ll,

corresponding to the orientation of the headpiece. Thirty

two samples (determinations of the coordinate values) were

made for each test measurement.

The DAS was first checked for resolution (output

consistency) with the headpiece positioned at a single

location. A similar approach, discussed in section D.6 in

the Appendix, was used in designing the simulation of the

stimulus jitter introduced by the variability of the raw

position data. The data distribution used in the

simulation, although not repeated here, can be considered as

additional characterization of the DAS.

With the CMTS adjusted to give nominal Y, X and

orientation coordinate positions of (0,0) mm and 00

respectively (graphically displayed in Figure 4), four tests

were conducted with the resulting data shown in Table 2.

Within the short time span of a single characterization

trial (3.2 sec), the standard deviations of the position

means primarily reflect experimental uncertainties and noise

in the system rather than drift. The means of these

standard deviations are referred to in the table as

intertrial means of intratrial standard deviations. In the

first test, 32 trials were performed in quick succession.

Even so, it is seen that the position means do drift

slightly as indicated by the larger standard deviations of

these means, which are referred to as intertrial standard

deviations of intratrial means. In the second test, the


CMTS was checked for confidence of position repeatability by

moving the headpiece and then returning it to the same

position (as indicated by the CMTS vernier dials) before

each of 32 trials. The results corroborate the accuracy of

the CMTS by showing a maximum standard deviation of position

means increase of 0.04 mm beyond that found in the first

test. The third test further investigated position data

fluctuations due to drift over 25 trials. In this test the

PSS was powered down for awhile after each of the first two

groups of 5 trials and then the PSS was allowed to idle with

power on for an hour after each of the second two groups of

5 trials. The results show a substantial increase in the

standard deviations of the position means. The fourth test

examined the influence of acoustic source use on drift and

involved transmitting 2000 clicks between each group of 5

trials for a total of 15 trials. The results of this test

indicate that acoustic source usage does not exert a strong

influence on drift.

The next series of tests characterizes the DAS for

headpiece position data accuracy over a portion of the

system's operational range with the use of the CMTS. Of

interest is not only the accuracy of the absolute coordinate

values but also the accuracy of changes in the coordinate

values resulting from small positional perturbations

(indicating relative position accuracy) over a range of

nominal positions. The results of 5 tests are summarized as

error means and standard deviations in Table 3. In each


test the absolute position data errors are calculated by

using the line which best fits the data points as found with

linear regression analysis. The errors found with lateral

(along the X axis) translation are somewhat affected by

headpiece orientation and are as much as twice those found

with diagonal (along the -450 line) and anteroposterior

(along the Y axis) translation. The results also indicate

that perhaps more than half of the magnitude of these errors

can be attributed to the DAS fixed position resolution

limitations previously presented.

Table 2
DAS Accuracy with a Stationary Subject

Parameter Y Axis X Axis Rotation
(mm) (mm) (deg)

Test 1 Rapid succession 32 trials
Intertrial mean of intratrial S.Ds. 0.25 0.32 0.07

Intertrial S.D. of intratrial means 0.27 0.43 0.08

Test 2 Subject moved and returned 32 trials
Intertrial mean of intratrial S.Ds. 0.24 0.33 0.07

Intertrial S.D. of intratrial means 0.31 0.45 0.08

Test 3 System turned off and on 25 trails
Intertrial mean of intratrial S.Ds. 0.29 0.33 0.08

Intertrial S.D. of intratrial means 0.54 0.50 0.12

Test 4 Extensive source operation 15 trials
Intertrial mean of intratrial S.Ds. 0.25 0.30 0.07

Intertrial S.D. of intratrial means 0.22 0.40 0.07


Table 3
DAS Operational Range Accuracy



along X axis
w/ 00 rot.

along X axis
w/ -430 rot.

along -450
w/ 00 rot.

along Y axis
w/ 00 rot.



-531 mm
0 mmb

-340 mm
0 mmb

-20 mm
435 mmc

-147 mm
312 mmc

Absolute Position
Err Mean Err S.D.
0.200 0.150

0.9 mm 0.7 mm

0.9 mm 0.5 mm

0.4 mm 0.2 mm

0.5 mm 0.2 mm

Relative Position
Err Mean Err S.D.
0.110 0.090

0.6 mm 0.4 mm

0.4 mm 0.4 mm

0.3 mm 0.3 mm

0.4 mm 0.4 mm

a) Data point intervals are 60 with 0.30
relative position measurements.
b) Data point intervals are 38.1 mm with
for relative position measurements.
c) Data point intervals are 57.2 mm with
for relative position measurements.

perturbations for

2 mm perturbations

2 mm perturbations

The speed of the DAS, alluded to in the software

description (section 2.1.3), and other factors affecting the

performance of this implementation, are discussed among

conclusions regarding the PSS in section 5.1.1.


2.2 Stimulus Control Subsystem

The Stimulus Control Subsystem (SCS) is designed to

generate analog waveforms that drive the Stimulus Projection

Subsystem. The SCS is predominantly based on software and,

in the implementation discussed here, performs extensive

machine language level processing in generating these

waveforms as a function of the raw position data it receives

from the DAS.

2.2.1 SCS Hardware

The SCS hardware is composed of the stimulus

microcomputer, an interface to the parallel link, and a

12 bit digital to analog converter. Its function, in

conjunction with appropriate SCS software, is to generate

analog waveforms that control the movement of the visual

stimulus. In this particular study, the major task of the

SCS is to receive raw position (timer) data from the DAS,

process the data, and update the stimulus position, all as

quickly as possible in order to minimize the response delay

between the subject's movement and stimulus movement. In

order to achieve this, the stimulus microcomputer, an Ohio

Scientific Challenger IP microcomputer similar to the

position microcomputer, was made operational at 2 MHz by

swapping in faster chips and rewiring the clock inputs. A


6522 VIA chip was installed to provide ports for the D/A

converter and the parallel link. The D/A converter,

discussed in detail in section C in the Appendix, is based

on a NS DAC1210LCD chip, and control signals are derived

specifically to enable it to perform at maximum speed in

conjunction with the SCS program operating at a machine

language level. In addition, the interrupt handler of the

computer's operating system, which was necessary to receive

the raw position data with greatest time efficiency,

suffered from a design error which had to be corrected (see

section D.1 in the Appendix).

2.2.2 SCS Geometric Analysis

One of the main functions of the SCS is to calculate

the desired stimulus position based on the position and

orientation of the subject's head and the selected feedback

gain of a trial. This calculation, performed by the SCS

software (section 2.2.3), involves the use of ten equations

that are discussed in this section.

The first four equations are used to calculate the X

and Y coordinates of Sp 1 and Sp 2 which define the position

and orientation of the subject's head (see Figure 4 and

discussion in section 2.1.2). These equations are of the

form shown in equation (5), section 2.1.2, except that the

units of the coordinate values are changed to radss" in


order to simplify the remaining equations. A rad is defined

as a length equal to the radius of the cylindrical stimulus

screen and, although designed to be 80 cm, has been

empirically found to be 77 cm (see section 2.3.2). The

generalized expression of these four equations for a

coordinate value in rads is then

(6) CV = ((CA)2 (CB)2 + 351(CB CA))/12498510

where CA and Cg are sp-mic pair raw position data that are

picked by the SCS software based on its error checking and

correcting functions. See Table 1, section 2.1.2.

The next four equations are used to calculate the

stimulus position for the normalized case of unity feedback

gain. This case corresponds to the stimulus moving with the

subject's head so that it appears stationary to the subject.

Referring to Figure 4, section 2.1.2, the intent is to

calculate the stimulus angle 8 such that the stimulus

projection line (L3) intersects the subject's line of sight

(a segment of line L2) at the surface of the screen'. A

subject standing at the center of the screen (on the axis of

rotation) and facing straight ahead (along the Y axis in its

positive direction) would result in e = 0. The equation of

the line that includes the subject's line of sight (L2) is

defined as

(7) X = mY + b

meaning that m = 0 corresponds to the subject facing

straight ahead (parallel to the Y axis). Line L2 is assumed


to be orthogonal to the line connecting the acoustic sources

(LI), and thus the slope of line L2 is

(8) m = (Y2 Y1)/(X2 Xl)

Line L2 bisects line Ll at a point whose (Y,X) coordinates

are ((Yl + Y2)/2,(X1 + X2)/2), so from equation (7) the X

intercept of line L2 is

(9) b = (Xl + X2 m(Yl + Y2))/2

The equation describing the screen is

(10) X2 + y2 = 1

since the screen radius is defined as 1 rad. The X

coordinate of the intersection of line L2 with the screen,

from equations (7) and (10), is then

(11) Xp = b + mYp = (1 (Yp)2)

Squaring both sides of equation (11) and solving the

resulting quadratic equation for Yp, the Y coordinate of the

intersection of line L2 with the screen is

(12) Yp = (- bm i (m2 b2 + 1))/(m2 + 1)

There are two values for Yp corresponding to two possible

intersections of line L2 with the screen. Since it is known

that the subject's line of sight is always in the positive

direction with regard to the Y axis, the greater value of Yp

is the only one of interest. In addition, if the radical in

equation (12) is defined as

(13) S = (m2 b2 + 1)

then equation (12) becomes

(14) Yp = (- bm + S)/(m2 + 1)


The tangent of the normalized stimulus angle, with the use

of equation (11), is

(15) tan en = Xp/Yp = (b + mYp)/Yp

and finally, using equation (14) in equation (15) and

simplifying, the normalized stimulus angle is

(16) en = arctan ((b + mS)/(S bm))

The following two equations are involved with initial

position correction and trial feedback gain. The stimulus

angle is initialized to 0 at the beginning of a trial to

allow for the greatest possible stimulus movement in either

direction before reaching the physical limits of the SPS

(resulting in nonlinear "clipping"). The initial position

and orientation of the subject might not be in absolute

accord with the initial stimulus angle, and so as to prevent

an abrupt initial stimulus movement at the beginning of a

trial, these initial positions of subject and apparatus are

coupled by setting an initial position correction offset.

This also compensates for any malposition of the acoustic

sources on the subject's head (i.e., subject's line of sight

not being exactly orthogonal to, or bisecting, the line

connecting the acoustic sources). This offset is then used

in calculating each new stimulus position update in order to

maintain smooth stimulus motion during the remainder of the

trial. Thus the relative movement between the subject and

the stimulus is maintained, although the absolute position

of the stimulus is not solely determined by the present

subject position. This is of no consequence in this study


where the stimulus is a vertical grid pattern covering the

full field of vision of the subject. The equation for the

initial position correction offset is

(17) C = 8i Gen

where 8i is the initial stimulus position, and G is the

trial feedback gain condition. This equation is only used

once during each trial with the first valid value of 8n

calculated by equation (16). Once the value of this offset

is established, then the equation for the actual stimulus

angle that is used throughout the remainder of the trial is

(18) ea = Gen + C

This actual stimulus angle is the final result that

needs to be sent to the 12 bit D/A converter for controlling

the SPS. Before the result can be used however, it needs to

be modified for compatibility with the converter. For the

sake of time efficiency, this preconditioning of the actual

stimulus angle is incorporated into equations (17) and (18)

as they are used in the SCS software (see section 2.2.3).

2.2.3 SCS Software Background. The SCS software, developed for

this study, is an online system which manages the

utilization of the received raw position data to generate

analog waveforms which are user specified functions of this

data. This software, which is the heart of the SCS, is


involved with (a) the generation of a random sequence of

trial stimulus conditions, (b) the initialization of the

mass storage device for cataloging these trials into stored

data files, (c) the task of reception, error detection and

possible correction of the raw position data sent by the

DAS, (d) the calculation of the position and orientation of

a subject from each complete raw data sample, (e) the

calculation of the desired stimulus position based on the

subject position and orientation result and the trial

stimulus condition gain or (f) the calculation of the

desired stimulus position based on a distribution simulating

the jitter introduced by the DAS's experimental

uncertainties and the trial stimulus condition gain, and

(g) the conversion of the desired stimulus position to an

analog form by interfacing with and generating waveforms

used for the control of a D/A converter.

The major concerns in the development of this software

were speed and size. The stimulus microcomputer only has 8K

of user available RAM to perform all the above functions.

Speed is of the utmost importance since this determines how

long it takes for the most recently determined position of

the subject to be reflected in an updated position of the

stimulus. In order to maximize the speed of the software,

all the online functions during a trial are performed at a

machine language level. In addition, a number of unorthodox

practices are incorporated which, while aiding with speed

and size, make the code difficult to follow and to modify.


The more complex equations, used in the calculations of the

subject position and orientation and desired stimulus

position, are evaluated using BASIC floating point variables

and the available operating system ROM routines, but all

accessed from the machine language level. The number of

BASIC variables used is minimized by their multiple usage to

reduce allotted variable array space and access time;

however, this makes the software maintenance difficult. In

addition, some numerical constants are represented by

variables to improve access speed, and equations are

arranged with regard for the speed of different algebraic

operations, all in order to reduce computation time. The

machine language portion of the software consists of six

routines imbedded in a BASIC program which performs offline

tasks and acts as a user interface. The complete program

with allocated memory uses all but about 100 bytes of the

available RAM. Therefore, the BASIC portion has virtually

no documentation included together with the code. An

operational overview of the software is presented in the

following sections, and the actual code with separate

documentation and memory map of the system is presented in

section D in the Appendix. Pretrial operation. Operation of the SCS

software begins with the BASIC program, DSTIM.BAS,

generating a random sequence of trial conditions. There are

14 different trial stimulus conditions corresponding to


(a) positive and negative feedback plus baseline conditions

for 4 different nonzero values of stimulus gain, and

(b) both an eyes opened and eyes closed condition for zero

stimulus gain (stationary stimulus). The eyes opened zero

gain condition is picked twice in the generation of 15

randomly sequenced trials. The entire 15 trial sequence is

repeated in reverse order (counter-balanced) for a total of

30 trials per subject run. The sequence is selected with

the computer's pseudo-random number generator and begins

with the (N x 15)th random number where N is the subject

number (1 to 33). At the user's command, the subject's

trial list is printed as hardcopy and then downloaded to the

DPS, to organize the filing of the forthcoming raw position

data. DSTIM.BAS then pokes (loads) memory with initial

values for buffers, code for the machine language routines,

and algebraic equations in BASIC's intermediate code or

"tokenized" form. The interface for the parallel link is

initialized to effect an interrupt of the SCS when -the DAS

is ready to transmit data, to notify the DAS when it is

ready to receive each byte, and to latch each byte when the

DAS indicates data ready. The D/A converter interface is

initialized for sending data and for generating waveforms

involved with the production of control signals for the D/A

converter (see section C.1 in the Appendix). Some status

and error flags involved with the operation of the machine

language routines are now cleared offline. The mic error

criteria values for online checking of the raw position data


(discussed later in this section) are entered from DSTIM.BAS

and poked into machine language code in order to simplify

access. Information sent to the D/A converter initially

centers the visual stimulus on the screen, and some

numerical constants are assigned to BASIC floating point

variables. The test condition for the first trial is

displayed on the monitor along with "STIMULUS ERRORS= 0" and

a message stating how to enter a different trial. Every

occurrence of a stimulus error updates both the associated

counter and the displayed message. The option to enter a

different trial can be invoked by pressing the key at

any point of operation to stop the SCS software. Then, by

performing a warm start ( key), hitting any key and

, and entering the command "GOT0999", DSTIM.BAS allows

the user to resume the program at any requested trial in the


After the monitor display is updated, the interrupt

capability of the stimulus microcomputer is enabled and the

software begins online operation. If the trial is a

feedback condition, then the trial gain is preconditioned by

a numerical factor to achieve compatibility with the D/A

converter (as discussed in section as assigned to

the variable G, and the machine language routine, ERROR.MLP,

is entered. If the trial is a baseline condition, a

DSTIM.BAS routine is entered which simulates the stimulus

position jitter induced by the DAS (discussed in section

78 Feedback condition trial. An operational

flowchart of the machine language routines is shown in

Figure 6. It begins with ERROR.MLP being either (a) just

entered from DSTIM.BAS, or (b) looped back to after

processing the previous data. After performing some

initialization tasks, this routine checks the data error

flag for a data use error, which occurs when a new sample

data from the DAS arrives before the new data buffer is free

(an overflow). This error indicates that the DAS sampling

rate is too fast for the SCS to keep up with, and results in

(a) the SCS returns control to DSTIM.BAS, (b) an error

message is displayed, and (c) the SCS program ends. If no

data use error occurred then ERROR.MLP checks the data-in

flag for the arrival of, and if necessary waits for, the new

position sample raw data transferred from the DAS.

The DAS transmits position data grouped by acoustic

source, and each transmission generates an interrupt that

switches control to the interrupt handler, DYNHAND." First,

DYNHAND checks the data-in flag to check whether the new

data buffer is free, and if it is not, the data error flag

is set and the handler returns from the interrupt. If the

buffer is free, DYNHAND stores the incoming data group in

its appropriate location in the new data buffer and then

returns from the interrupt. This buffer can contain a

complete raw position data sample (all the data groups) and

DYNHAND only sets the data-in flag when the buffer is full.

Enter from BASIC
w/desired Gain


( rInterrupt by
Position Computer (C2)
................. ...................................

Previous N Set Data Use
Data Used? Error Flag

Receive -"
New Data -" -
Clc S u ." P

SReturn from Interrupt) -s G
... .. .............. ................ Loop fo



Var.MLP Assign to BASIC Variables
& Declare Data Used

EqRun.MLP. Calculate Stimulus Position
busing ROM Routines & Gain

................... ................ '. Error Counter

Position in Nh Return to BASIC :
Bounds' w/Error Message '

Update D/A
... ............................... ....
Figure 6. SCS Software Flowchart: Online Operation for
Stimulus Feedback Trials. Raw positional data from the
DAS are error processed and used in the calculation of a
variable gain visual feedback signal.


Nect Pr Data Use
cesError &

New Data N
i Return to BASIC
Y w/Error Message
r eachStim

Datum Pass Min.N N Set Pair
& Max. Criteria? 'Error Flag

Datum Pass N Set Pair
Window Criteria? Error Flag

pdate Pair Datum

Y Another

ect Pair Data using
ces & Pair Error Flags

Data N

Termination Y Return to
Code), BASIC /

Increment Stimulus


Upon determining that a new data sample has been stored

in the new data buffer, ERROR.MLP begins error checking.

The geometric relationships between acoustic sources and

microphone/detectors, as depicted in Figure 4

(section 2.1.2), are referred to here. Each of the sp-mic

pair data in the new data buffer, corresponding to one of

the eight distances Dl through D8, is checked for meeting

three criteria (whose test values were entered via

DSTIM.BAS). Two criteria are minimum and maximum allowable

values for the datum, each associated with the operational

limits for these distances (i.e., the limits of the

measurement volume). The third criterion is a window test

based on the fact that an acoustic source can only move a

finite distance between successive samples.

One problem of a window test is in allowing the

collection of valid data to "get started" from an initial

situation of no pair data (all initial datum values are

$0000). Another problem is to "catch up" after one or more

datum errors of a particular sp-mic pair since by then the

acoustic source might have moved away from a previous

position not to return to that vicinity again. The solution

is a window comparison of a sp-mic pair datum with the last

datum of that same pair which had passed the minimum and

maximum criterion tests (but not necessarily the window

test). This is implemented by having an old data buffer,

structured like the new data buffer, which stores the sp-mic

pair data to be used in the next collected data sample's


window criterion tests. Whenever a sp-mic pair datum in the

new data buffer fails either the minimum or maximum

criterion tests, an error flag for that pair is set and the

datum presently in the old data buffer for that pair is left

unchanged. However, if these two criterion tests are

passed, but the window criterion test is not, the error flag

for that sp-mic pair is still set but, as when all the tests

are passed, the new pair datum overwrites the datum in the

old data buffer.

ERROR.MLP performs its error correcting function by

evaluating the results of these criterion tests and deciding

whether to, and with which sp-mic pair data, effect a

stimulus position update. The sp-mic error flags permit

judgements as to whether each of the four coordinate values

(two for each acoustic source) can be calculated using

(a) the default, or if not, (b) the alternative coordinate

determinants, or if neither then (c) the stimulus position

is not updated. See Table 1 in section 2.1.2 for default

and alternative coordinate determinants.

If all four acoustic source coordinate values cannot be

calculated, a flag is set and control is passed to TERM.MLP.

This routine determines why the stimulus position update

could not be accomplished. The possible reasons are

(a) true datum errors, (b) every datum in the first sample

of a trial always fails the window test, or (c) the

termination code was sent by the DAS signifying the end of a

trial. In the first case, the STIMULUS ERRORS counter is


incremented, the data-in flag is cleared indicating that the

new data buffer is free, and the program returns to

ERROR.MLP to await the arrival of the next position data

sample. The second case is treated similarly except that

the STIMULUS ERRORS counter is not incremented. In the

third case, further interrupts are disabled and the program

returns to DSTIM.BAS which then reports the end of the trial

and prepares for the next trial in the sequence.

If all four acoustic source coordinate values can be

calculated then the sp-mic pair data are arranged in

appropriate locations in the new data buffer, which now

serves as an "equation variables buffer", to be assigned to

BASIC variables by a routine called VAR.MLP. This buffer is

reorganized in a manner which capitalizes on the existing

positions of the pair data within it in order to minimize

the amount of data movement and time expenditure necessary

(as shown in section D.4 in the Appendix). In order to

avoid any data loss due to overwriting in the equation

variables buffer, any sp-mic pair data that needs to be

moved is read from the old data buffer (which by now

contains the pair data from the present sample) into this


VAR.MLP now assigns each of eight BASIC variables

(S through Z) to a different 2 byte datum in the 16 byte

equation variables buffer, corresponding to the eight

distances chosen to be used in the calculation of the

acoustic sources' coordinates. This is accomplished by


using (a) an operating system ROM routine which converts

each 2 byte datum to a floating point number, (b) another

ROM routine which creates BASIC floating point variables,

and (c) a third ROM routine that sets pointers which equate

the variables to the numbers. With all the BASIC variables

assigned, the data-in flag is cleared to indicate that the

equation variables (or new data) buffer is now free.

The process continues with a routine, EQRUN.MLP,

directing calculations using the algebraic equations

previously loaded into memory in BASIC "tokenized" form by

DSTIM.BAS. A pointer is set to the location of each

equation which is then evaluated by a ROM routine. The

first four equations, used to calculate the four coordinate

values, are in the form of equation (5) in section 2.1.2 and

in BASIC nomenclature are

(19) S = (T*T-S*S+P*(S-T))/Q

(20) T = (V*V-U*U+P*(U-V))/Q

(21) U = (Y*Y-Z*Z+P*(Z-Y))/Q

(22) V = (X*X-W*W+P*(W-X))/Q

where the variables S through Z on the right sides of the

equations were assigned by VAR.MLP as described above and

the variables S, T, U, and V on the left sides of the

equations correspond to the coordinates Y1, Xl, Y2, and X2

respectively. As previously described, the multiple use of

BASIC variables saves memory and execution time, as does the

multiplication of a variable by itself rather than squaring

it, and also the use of the variables P and Q rather than


numerical constants. DSTIM.BAS previously assigned the

values P = 351 and Q = 12498510 offline, with Q being

different than in equation (5) in order to have the units of

the coordinates come out in rads (1 rad = 770 mm).

The next four equations, used in calculating the

normalized stimulus angle, are taken from equations (8),

(9), (13), and (16) in section 2.2.2 and in BASIC

nomenclature are respectively

(23) W = (S-U)/(V-T)

(24) X = (T+V-W*(S+U))/A

(25) Y = SQR(W*W-X*X+N)

(26) T = ATN((X+W*Y)/(Y-X*W))

Again, the definitions of variables might change in each

equation, but they can be ascertained by comparing these

equations with those they are taken from. The two

variables, A and N, were equated to the numerical constants,

2 and1. respectively, by DSTIM.BAS. The normalized stimulus

angle, T, is expressed in units of radians by the arctan ROM

routine, ATN.

The last two equations, used in calculating the initial

position correction offset and actual stimulus angle, are

taken from equations (17) and (18) in section 2.2.2 and in

BASIC nomenclature are respectively

(27) C = C-G*T

(28) T = G*T+C

When EQRUN.MLP is entered for the first time in a given


trial, equation (27) is used to calculate the initial

position correction offset, C (on the left side of

equation (27)), and the initialization flag is set so that

this equation is bypassed in subsequent usage of this

routine. The actual stimulus angle, T (on the left side of

equation (28)), is the result to be sent to the 12 bit D/A

converter for controlling the SPS. Incorporated into the

calculation of T is a preconditioning for compatibility with

the D/A converter. The angle sent to the converter first

needs to be converted to an integer in the range of 0 to

4095 (a 12 bit binary number) corresponding to a stimulus

movement range of 0 to 320 (as dictated by the limitations

of the SPS as discussed in section 2.3.2). It should be

noted that the preconditioning includes a change of

convention for the stimulus angle from a straight ahead

position (along the Y axis in Figure 4) being called 0 to

being called 160. After the angle is converted to an

integer, it needs to be left justified 4 bit positions so

that its most significant bit is the most significant bit of

a 2 byte word (as discussed in section C.1 in the Appendix).

Two of these left shifts are later performed by a D/A

converter handling routine, D/A.MLP, as part of its out of

bounds error detecting function, and so only two left shifts

are incorporated into this present calculation of T.

For any stimulus angle within the acceptable 160 16

range, the calculation of the preconditioned actual stimulus

angle in equation (28) will result in the BASIC floating


point variable, T, having a value between 0 and 16380 ($0000

and $3FFC). The routine D/A.MLP begins by using a ROM

routine to convert the angle T to a 2's complement 16 bit

integer. In this form, an angle within the acceptable range

has its two most significant bits being Os since the

preconditioning only included two of the four left shifts

needed for full left justification of the original 12 bit

number. This allows for quick testing of an angle to

determine if it is within the bounds imposed by the SPS,

which in turn is now reflected in the input capability of

the D/A converter. D/A.MLP performs a left shift on the

16 bit integer, and if the overflow bit is a 1, then the

angle is less than 0 (a negative value) and the stimulus

update is aborted. If the angle is not less than 0, then

another left shift is performed, and if the overflow bit is

a 1, then the angle is greater than 320 (320 would now be

represented by the value $FFFO) and the stimulus update is

aborted. In the event of an abort which ends the trial, the

error and clipping flags are set, further interrupts are

disabled, and control returns to DSTIM.BAS. For the

convenience of the user, DSTIM.BAS provides the option of

either repeating the trial or going on to the next trial.

If the angle was within the acceptable range, then

D/A.MLP transfers the 16 bit number, a byte at a time along

with the appropriate control signal generating bytes (see

section C.1 in the Appendix), to the D/A converter. A pause

during the transfer, necessary because of the converter's


operating and settling times, is efficiently used to perform

some initialization for the forthcoming next operation of

ERROR.MLP. Upon completion of the transfer, which results

in the stimulus position update, the program returns to a

postinitialization entry point in ERROR.MLP for processing

the next position sample data in the trial.

Preconditioning for D/A conversion is accomplished by

the utilization of numerical constants in the equation

variables as follows. An angle expressed in radians can be

converted to the proper form by multiplying it by the factor

J where

J = 180/pi (degrees/radian) x 4096/32 (digital levels/degree)

x 4 (effects two left shifts of a binary number)

= 29335.439

In equation (28), the value of T on the left side would then

be in the proper form if C (the initial position correction

factor) already was in that form, and the term G*T was

multiplied by J. This C (the initial position correction

factor) would be in the proper form if in equation (27), C

on the right side (the initial stimulus position) already

was in that form, and the term G*T was multiplied by J.

Since the stimulus is initially centered on the screen by

DSTIM.BAS, the initial stimulus position, C, in the proper

form, is 160 x pi/180 (radians/degree) x J = 8192. This

numerical value for C is assigned by DSTIM.BAS before the

beginning of the trial. The multiplication of the term G*T

by J in both equations (27) and (28) is accomplished (as was


mentioned in section offline by DSTIM.BAS

multiplying the trial gain G by the factor J just before

ERROR.MLP is entered. Baseline condition trial. For a baseline

condition trial, DSTIM.BAS uses a routine which simulates

the stimulus position jitter induced by the DAS. This

jitter is a result of noise and of fluctuations in the

characteristics of the DAS components, and because of the

quantizing effect of the DAS timers, it appears as a

discrete variation in the stimulus position. A discussion

of the design of the simulation, and a justification of its

validity, appears in section D.6 in the Appendix. The

routine begins by waiting for a complete raw position data

sample to be received by the SCS (as indicated by the

setting of the data-in flag which it then clears), after

which a time-out counter is cleared and a random number

between 0 and 1000 is selected. This number is used to

access a weighted tree which returns a normalized integer

stimulus rotation angle from -7 to +7. These integers, when

multiplied by 0.0434, yields stimulus position angles such

as are observed with the jitter introduced when unity gain

feedback is employed and a stationary headpiece is

positioned at the center of the SPS. The preconditioned

actual stimulus angle is calculated from

(29) Y = 32768+C(0,T)*Y*89

Full Text
xml record header identifier 2009-01-26setSpec [UFDC_OAI_SET]metadata oai_dc:dc xmlns:oai_dc http:www.openarchives.orgOAI2.0oai_dc xmlns:dc http:purl.orgdcelements1.1 xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.openarchives.orgOAI2.0oai_dc.xsd dc:title Acoustically measured postural stability with visual feedbackdc:creator Shuman, Dennisdc:publisher Dennis Shumandc:date 1987dc:type Bookdc:identifier (oclc)000943148 (alephbibnum)dc:source University of Florida