Video Tracking fot Gait Analysis

8/15/2019 Video Tracking fot Gait Analysis

1/52


2/52

REHAB Tech- Monash Rehabilitation Technology Research Unit assume no liability for any

claim of adverse effects resulting from misapplication of the information presented here in.While every effort is made to ensure the accuracy of the guide no responsibility or liabilitywill be taken for any inaccuracies.

REHABTech is finance and supported by

In collaboration with

Copyright 1998

All rights reserved. No part of this publication may be reproduced or transmitted in any form

or by any means, electronic or mechanical, including photocopy, recording or any information

storage and retrieval system, without permission in writing from the publisher.

Requests for permission to make copies of any part of the work should be addressed to:

REHAB Tech- Monash Rehabilitation Technology Research Unit

C/- C.G.M.C.

260 - 294 Kooyong Road

CAULFIELD VIC 3162

AUSTRALIA

Email [email protected]


3/52

CONTENTS:

CHAPTER 1:

1.1 INTRODUCTION

CHAPTER 2:

2.1 AN INTRODUCTION TO GAIT ANALYSIS

CHAPTER 3:

3.1 OUTLINE OF TOTAL DESIGN

3.2 STRATEGIES AND IDEAS

3.3 ASSESSABLE SECTIONS

3.4 OTHER POSSIBILITIES

CHAPTER 4:

4.1 VIDEO CAMERA SELECTION

4.2 CAMERA/LENS OPERATION

CHAPTER 5:

5.1 CIRCUIT DESCRIPTION 5.2 VIDEO SIGNAL EXTRACTION

5.2.1 PAL VIDEO FORMAT

5.3 SCANNING LIMITS AND VIDEO OUTPUT

5.4 TIMING AND ANALOGUE TO DIGITAL

CONVERSION

5.5 MOTOR PULSE GENERATION AND SAFETY

5.6 LIMIT SAFETY FOR TROLLEY

5.7 AUXILIARY CIRCUITS

5.7.1 CRYSTAL OSCILLATOR

5.7.2 LED POSITION DISPLAY

CHAPTER 6:

6.1 ULTRASONIC DISTANCE MEASUREMENT

6.2 BASIC CONCEPT

6.3 CIRCUIT DESCRIPTION

CHAPTER 7:

7.1 MOTOR SELECTION AND CONTROL

7.2 MOTOR SELECTION

7.3 MOTOR DRIVING TECHNIQUES

CHAPTER 8:

8.1 CONSTRUCTION

CHAPTER 9:9.1 CONCLUSION

CHAPTER 10:

10.1 UPDATE 18.1.1996

BIBLIOGRAPHY


4/52

CHAPTER 1:

1.1 INTRODUCTION

Bipedal walking, the act of human locomotion and one of the definingcharacteristics which separates us from the rest of the animal kingdom, has been a

subject of study in one way or another ever since mankind first performed the actitself. The high efficiency with which humans can undertake what amounts to be anincredibly complicated mechanical function is reason enough in itself to stimulate

study. To appreciate the complexity of the task one must only observe some of thehighly advanced, yet still unwieldy attempts that scientists and engineers have made toconstruct two legged walking machines. Still, researchers in every field find a

multitude of reasons to delve deeper into the topic and add to the already enormousvolume of data held on the subject. As a few randomly selected examples;

Prosthetists and Orthotists may want to examine a patients gait to improve their

walking style for greater comfort and function.

A biomedical engineer may study gait so as to construct a muscle stimulator toactivate legs left motionless by spinal damage

A chiropractor may study gait to find the source of a patients back pain.

Sports analysts may study a sprinters style to find a faster and more efficient wayfor the athlete to run.

Gait analysis could be a much simpler science if there existed a definable normalgait, but unfortunately there is no such concept. Every human on earth has a slightlydifferent way of walking. The factors which effect our walking are numerous; a

persons height, weight, sex, ancestry, mood, state of health and footwear just to namea few. The lack of correlation does not, however, render all data on gait assessmentuseless even if one style cannot be labelled as average. Patterns still emerge which

are common to all gait styles within subgroups, and models can be produced that fitgiven percentiles of the population. An orthotist does not fit a patient with anapparatus which will conform them to a normal style of walking, but rather to a gait

which is most efficient and comfortable for that individual, and which will not causefurther problems in the future. Trying to correct their gait to a theoretical normal stylecould cause greater damage.

What this project aims to produce is a tool which will enable professionals toanalyse an individuals walking style in a way not used before. The first questionwhich must be examined isHow does one investigate and record someones gait?There have been a plethora of qualitative and quantitative inventions to do this since

the early 1800s. Complicated stills cameras, interrupted light techniques, insertedpins, motion film, force plates, pneumatic shoes, glass walkways and

electromyographs have all been used, and many more. This project is using a videotechnique for qualitatively recording an individuals gait during a sequence of stepcycles.

Figure 1.1: Example of wireframe plotting Program

(Rose P216)


5/52

CHAPTER 2:

2.1 AN INTRODUCTION TO GAIT ANALYSIS

One of the first (recorded) observations regarding the science of human

movement was made by Aristotlearound 350BC. He noted that if a man were to

walk...alongside a wall (with a reed dipped in ink attached to his head) the line would

not be straight but zigzag, because it is lower when he bends and higher when he

stands upright and raises himself [17] . He went on to describe the transfer of body

weight from side to side during walking, and joint flexion and extension. At the time

his observations remained just that, but later work in the area tested Aristotles

theories on the subject, and found them to be accurate.

The next notable advancement in human motion analysis occurred in the

1600s. Around the time that Galileo was formulating his laws of motion, and Newton

was gathering inspiration for his most noted work, Principa, which described the basis

of modern dynamics, Giovanni Borelli[17] was applying these concepts in amechanical framework that viewed bones as levers and muscles as forces on those

levers. He also defined the location of the centre of gravity for humans. More

importantly, at this time it was now possible to test such theories experimentally.

Around two hundred years later Wilhelm and Eduard Weber[17] combined

their talents (which covered most of the known sciences) to produce some models of

physiological motion that were widely acclaimed in the field. Using simple tools

available at the time (distance measurement instruments and a timepiece) they made

observations such as head displacements, stride lengths, and dimensions of the limbs

in differing degrees of flexion. But more importantly than these measurements were

the complex deductions they made about this data, resulting in a detailed model of

mechanical locomotion. It must be noted, however, that because the available

instruments were limited, many of the Weber brothers theories were not based on

direct measurements and were later proved to be erroneous in some cases.

In the mid 1800s physiologists began to construct more complicated

equipment. A major jump forward in technology took place when French physiologist

Etienne Mareyand student Gaston Carlet[1,5,17] developed a method of obtaining

measurements from the subject

without major interference from

the scientists. They used a special

pair of shoes which had air

chambers in the mid metatarsalregion, and thus changes in the

pressure were able to be recorded

on an instrument carried by the

subject. Later, Carlet added extra

air chambers to the shoes and

confined the subject to walk in a

circle by using a rotating arm

which would hold the pneumatic

tubes for recording, and also avail

measurements of vertical and horizontal oscillations from any point on the body.

Figure 2.1: Carlets device for recording gait using

pneumatic soles and rotating arm (Gamble P205)


6/52

The photographic gait quantifying methods so common today had (as commonwith many scientific advancements) an unexpected origin. Late in the 1800s the then

governor of California had a bet with a friend that all four feet of a horse were, at

some instant, off the ground simultaneously. To prove the bet Eadweard Muybridge

[1,17] was hired to gather evidence. Muybridge set up a series of electrical switches

which would activate sequentially the shutters in a linear array of cameras, placed atthe side of a race track. The resulting photographs won the governor the bet, and

Muybridge went on to publish many studies of locomotion using his photographic

methods.

Marey seized upon this idea and constructed a camera capable of exposing the

subject on the same photographic plate many times, rather than using multiple

cameras. The special camera rotated the unique cylindrical plate so that sequential

images were exposed evenly around the outside of the plate.

Problems arose when human walking was to be investigated in detail since the

frequency of images needed caused the exposures in current methods to overlap.

Marey fixed this by only photographing essential parts of the body, and dressing the

subject in dark cloth with white stripes on regions of interest. The resulting procedure

provided Marey with plates which showed the succession of white stripes but inwhich the subject was not visible. Thus a pattern of sequenced movements was

recorded. This procedure was termed geometric chronophotography. Although this

data was still relatively imprecise, Marey was able to use it to make a basic definition

of the relationship between walking speed and energy expenditure.

German born Wilhelm Brauneand Otto Fischer [17], his student, wished to

perform an analysis of human movement by modelling the body as a connection of 12

linked segments. Using cadavers they found the centre of gravity of each chosen body

segment, but to complete the study experimental data of the displacement, velocity

and acceleration of each segment whilst the body was in motion was required. Thus

the pair needed a method to obtain instantaneous spatial measurements in three

dimensions. Since the methods used in the past by Muybridge and Marey yielded onlytwo dimensional information, Braune and Fischer modified the geometric

chronophotography method yet again, and set up a laboratory that used a new

stereometricprocess. Their version of stereometry used four cameras, two on each

side of the subject so that each point on the body was viewable from two perspectives.

Using this process each pair of two dimensional points could be transformed to one

three dimensional point, and finally the body could be tracked through three

dimensional space.

This is not as simple as it sounds however; Braune and Fischer made a special

plotting table to read the points from the photographic prints and coordinates were

recorded by hand. The calculations which resulted took many months to complete.The preparation of the subject was just as important for good results. The volunteer

Figure 2.2: An example of Mareys chronophotography

method (Rose P206)


7/52

was dressed in black and Geissler tubes were used for illumination points that wouldshow up easily on the plates. Geissler tubes consisted of rarefied nitrogen in a smallglass cylinder. Current passing through the gas caused it to illuminate and thus thepoints of light could be strobed electronically to obtain discrete measurements fromthe resulting prints. Black tape was used to secure these to the body, with only the

middle of the tube left uncovered, signifying theanatomical position to be tracked. This laboriouspreparation of the subject could take up to 8hours. Fortunately Braune and Fischer gainedmuch from this work. They were able tocalculate the length of each body segment, thecentre of mass of the body and velocity andacceleration of each joint. Using moments ofinertia of body parts (gained from previouscadaver studies) and calculations of inertia ofcentre of mass of those parts, forces acting on the

joints could be computed. Furthermore, the dataon these forces enabled the pair to find muscleaction present on each segment.

At the University of California[17] in 1945, the Engineering and Medical

School wished to undertake an accurate study on movement, focussing especially onthe rotations in the transverse plane. In order to study pelvic and lower extremityrotations pins were surgically implanted into the pelvis, femur and tibia of subjects.Using 35mm film and interrupted light techniques in conjunction with two (strain

gauge) force plates the team was able to do the most comprehensive study of the time,and the data accrued was invaluable as abasis for comparisons of amputee andvarious other pathological gaits.

The basic ideas that governmodern photographic techniques remainthe same as those used in Mareys times,as do the problems that encumber theprocedures. Modern film and videomethods have the advantages of fasterframes rates and high quality recordingequipment. Thus with faster imagerates, more accurate data can be liftedfrom the media. It has been shown instudies [13,18,19] that for a study of walking, 25 samples per second is sufficient foraccurate representation, with little difference from data obtained at double this rate.However, this project endeavours to produce a visual record rather than amathematical one, and for human vision 60Hz is a much more desirable scanning rate.This is exactly what video cameras use, so this media can be utilised safely for theproject at normal scanning rates. For the recording of running motion, over 200frames/sec should be used [19], but in all cases the more frames per second the better.

However it must be kept in mind that the data processing overhead also increases withframe speed. The highly respected physiologist Verne Inman [17] used over 400

Figure 2.3: An early stroboscopic flashphotographic record.

This subject was also fitted with white

stripe down the leg for clarity.

(Ghista P140)

Figure 2.4: Example of interrupted light technique.

A rotating slotted disc was placed in from of the

camera and bulbs placed on parts of the lower

body (Gamble P13)


8/52

frames/sec in one early study of walking to achieve great accuracy in speed andacceleration values. Additionally, if a three dimensional study is being taken, then thesubject must be shot by at least two cameras simultaneously, although many more(four to six) are often used in practice.

More relevant to this project are the various film and video methods that have

been used in the past, and are still used in gait laboratories today. The earliest andsimplest method (but most time consuming) was first

used in 1972 by Sutherland and Hagey[17] with

16mm film at 50 frames/sec. Markers were placed onthe subject, a reference clock placed in the field ofview and a cartesian coordinate grid placed on a wallbehind the subject. The film was developed andprojected onto a special motion analyser, from whichthe markers points were recorded by hand. This

tedious recording system was soon replaced by a

computer with a crosshair cursor that was placedover the point and automatically recorded thecoordinates [9,13,17,18].

Video systems have the advantage of an image that is in electronic form, andthus is more suitable for automatic processing. Early video systems used passivemarkers (markers made from reflective material) with a light source directly behind orsurrounding the camera. Images stored on video tape allowed the operator to mark(using a computer) the points on the digitised image, and these can be stored andprocessed at will. More recent systems can extract the marking points with littlehuman intervention, such as the result shown below.

Each point is able to be labelled (L.wrist, R,ankle etc)and studied singularly,or with the other sets of points from different joints. These systems can sometimesencounter problems with some pathological gaits, where the motion of one markermay overlap another (such as the toe and ankle markers in the diagram), and thecomputer may incorrectly assign the point to the wrong joint. These systems usuallyplace stringent conditions on lighting and subject clothing. A popular method is theuse of infra red lighting and reflectors, along with the appropriate cameras speciallymanufactured to record images in the infra red frequency range. This removes theusual problem of ambient light and unwanted reflections from skin, shoes and soforth. The active marker method is also relatively popular. Active markers, most oftena group of Light Emitting Diodes (LEDs), flash one at a time in a prearranged order.

Thus the computer can tell what marker is being detected by using the time frame,since only one marker will be visible at a time. Another similar method is to use

Figure 2.5: Operator using a computer

to enter data points from a motion film.(Winter P26)

Figure 2.6: Example of Computer Gait analysis Program

(Rose P215)


9/52


10/52

CHAPTER 4

4.1 VIDEO CAMERA SELECTION

The choice of video camera is pivotal to many other parts of the project.

Besides the most important features such as size and weight, there are also optionsinvolving format, resolution, quality, compatibility and cost. The basic guidelines setbefore looking at any cameras were that the unit should satisfy these requirements:

Small

Lightweight. These two stipulations are to enable the pulley and motormechanism to operate as quickly as possible. The greater the weight and size,

the more powerful the motor, and the stronger the pulley mechanism will needto be.

Able to be affixed and detached to brackets easily. To allow easy setup and

dismantlement of the device.

Colour. Simply for viewing quality.

Composite Video output. RF output would require extra demodulationcircuitry, and since composite output is often available, this added complexity

may as well be avoided.

Low voltage supply. The camera will be several meters from a power source,and will be close to a large amount of sensitive circuitry, so mains power is

best avoided. The cables needed to supply low voltage power are also lighterand more flexible.

Able to be fitted with a zoom lens ,or any other lens that may be required.

Many cameras have fixed lenses. In the future this camera may be needed formany other projects at Rehab Tech, so this modularity will be useful.

Zoom and Focus on the lens MUST be manual. As the camera moves, thedistance from it to the subject will change often. Many cameras haveautomatic focus, motion compensation and zoom (using distancemeasurement transducers) and these have the effect of blurring rapid

movements. This should be avoided.The first cameras investigated were the two already owned by Rehab Tech. They wereunsuitable for the following reasons;

1. JVC GR570 This camera is relatively small and lightweight. It is termed ahandheld unit. Unfortunately the lens is fixed, and not capable of large zoomsor manual focussing. The zoom and focus are automatic and there is no

override.

2. JVC GF51000H This camera had suitable zoom capabilities, and theautomatic focus can be overridden. However the unit is large and heavy, as it

is mostly used for professional video results.Next some commercial, handheld video cameras were investigated. These had somegeneral problems, mostly due to market trends in this area, which made this style of

unit inappropriate.

Very few had external output sockets. Most had direct record to tape only.

Most had automatic zoom only. Manual zoom override was a rare option, and

alway added in excess of $600 to the price of a fully automatic unit.

Few had mounting holes.

All were expensive.


11/52

All models had some zoom capability, but this was always limited, and notsuitable for the projects requirements.

Thus investigation was focussed on the so-called CCD variety of camera. CCD

(Charge Coupled Device) actually refers to the sensing chip, which converts the

image into electronic signals. It is common now, however, to use the term CCDcamera when referring to small, self contained units, most commonly used in closedcircuit security situations. These cameras are, in general, small, lightweight, lowpower and low voltage, and able to take a large variety of lenses. Furthermore, for the

security market, a large array of lenses are available. A survey was conducted oflocally available CCD cameras, along with their defining characteristics, and these areoutlined in Table 1.

The columns are labelled as:

NAME: Brand name.

MODEL: Local model number

COL/BW:Is the camera colour or black and white.

POWER SUPPLY: External power supply needed by the camera to operate. (NOT

supply voltage of the internal circuitry)

COST incl tax: Using sales tax at 22%.

COST ex tax: Cost without sales tax as of 30/7/95

CCD SIZE: This is the actual surface area of the CCD sensing chip, which is locatedbehind the lens. The most common sizes are 1/2" and 1/3", where the smaller is a

more recent development. 1/4" CCD chips are currently coming onto the market. TheCCD size determines the fitting of the lens; a lens for a 1/2 camera will not fit onto a1/3" unit.

LINES OF RES: Number of lines of resolution on the CCD chip. Regardless of this

resolution, the output of their camera is still the standard 625 line format used inAustralian Television. The image from the CCD ship is digitally processed by the

camera and spread over the 625 lines.

VIDEO OUTPUT: Format available at the output socket. 1Vpp PALis most commonand refers to composite output. (Although 1Vpp is always stated on specifications, theactual maximum peak to peak voltage of the signal is 2.2V) Some cameras (Such as

the PULNIX) have two formats available. SVHS is SuperVHS, and transmits colourand image intensity separately. Composite format combines the two into one signal,and is the format to be used in this project.

SUPPLIER: Local supplier. This is not necessarily the only supplier of a particular

camera in the area; simply the ones that were visited.

The model decided on was the Jetcom colour camera, in the 1/2" size. This was

influenced by the fact that a suitable lens (to be discussed) was available through thesame supplier.

One of the initial requirements for the optical portion of the project was that

the lens be capable of manual zoom and focus. So there remains the quantitativedecision of lens choice. And as CCD cameras and lenses are sold separately, there islittle restriction. In the market place, there are four common ranges available in zoom

lenses for CCD cameras, and these are identified by their scope of focal lengths. Theyare

8-48mm

8.5-51mm


12/52

9-57mm

12.5-75mm

On detailed specifications more useful data is available, the most relevant here are the minimum object distance

angle of view (horizontal and vertical)

By the design in chapter 3, the subject will be no less than 2m away. The smallestregion that will want to be focused on will be a single joint (eg; knee, ankle, hip), andthe largest is the entire lower body. In graphical form, refer to the figure below.

From the diagram,

a = 0.5(vertical angle of view) assuming camera is oriented horizontally, and nottilted.h = height visible in image if object is 2m away.

so

ha

= 42

tan( )

If the lower body is approximated to be 1.2m in height, then the required angle is at

least

a = = 2 1 24

33 31tan ( . ) .

The minimum viewing angle is less vital, but if the smallest height required is set to

be 0.2m, then, from the formula above, the minimum required angle is 5.72. Thus a

zoom lens is required which covers the range of 5.72-33.3for the vertical viewing

angle. For a 1/2" camera, the 8-48mm zoom comes closest to covering this stipulationin most models. As a typical example, the eventual model chosen has a vertical angle

of view range of 5.22-33.1, which is satisfactorily close to the theoretical

requirement. The models with other focal length ranges may be able to fit more of thesubjects body in a wide angle, but not be able to zoom in as well, or alternatively be

able to zoom in well, but have insufficient wide angle capabilities, so the 8-48mmfocal length is the best alternative. Table 4.1 also outlines a range of available lenses.The columns are labelled as:

NAME: Brand name of lens.

RANGE: Focal length range in millimetres. AUTO indicates automatic iris control

(aperture control). MAN is manual aperture, as is required. Auto iris lenses are alsosignificantly more expensive.

COST: excluding tax only, as on 30/7/95

SUPPLIER: Local supplier name.

2ma h


13/52

CAMERAS

NAME MODEL COL/BW POWER COST COST CCD LINES VIDEO

SUPPLY incl tax ex tax SIZE OUTPU

ELMO COL 240V AC $900 $738 1/2" 370

ELMO BW 240V AC $600 $492 1/2" 370COMPUTAR FC 62B BW 24V AC $488 $400 1/3" 380 1Vpp PA

SAMSUNG CCS324AP COL 24V AC $915 $750 1/3" 320 1Vpp PA

JETCOM COL 12V AC $915 $750 1/2" 420

GOLDSTAR BW $400 $328 1/3" 400

JVC (CLONE) COL $560 $459 1/3" 330

PANASONIC WVCP412 COL 12V AC/DC $1,466 $1,202 1/3" 480 1Vpp PA

PANASONIC WVCP212 COL 12V AC/DC $1,069 $876 1/3" 330 1Vpp PA

PANASONIC WVBP104 BW 24V AC $830 $680 1/3" 380 1Vpp PA

PANASONIC WVBP312 BW 12V AC/DC $1,066 $874 1/3" 570 1Vpp PA

PULNIX TMC-63M COL 12V DC $1,878 $1,539 1/3" 450 PAL/SV

PULNIX TMC-6 COL 12V DC $1,854 $1,520 1/2" 450 PAL/SV

MINTRON OS 3511 COL 12V DC $1,011 $829 1/2" 380 1Vpp PA

PHILIPS COL 12V DC $1,318 $1,080 1/3" 1Vpp PA

LENSES

NAME RANGE COST SUPPLIER

KOSMIKA 8-48mm 600 ex tax COLLINS

KOWA 8.5-51 AUTO 959 ex tax PHILIPS

KOWA 8.5-51 MAN 456 ex tax PHILIPS

COMPUTAR 12.5-75 559 ex tax GEC

PANASONIC 8.5-51 AUTO 1356 ex tax GEC

D6x9.5AI 9.5-57 MAN 445 ex tax HADLAND

COMPUTAR 8-48 MAN 580 ex tax MAINLINECOMSICAR 8-48 MAN 580 ex tax MAINLINE

Table 4.1:


14/52

The final choice for the camera/lens pair is:

Jetcom colour CCD camera with 1/2" lens socket. ($750)

Computar 8-48mm zoom lens ($580) Also a 12VDC adaptor (unregulated) for supply to the camera.($20)

These were purchased from Mainline Security Products, Nepean Highway,

Elsterwick.

4.2 CAMERA/LENS OPERATION

The camera and lens are shown connected in figure 4.1. A VHS video tape isshown alongside for a dimensional reference. Table 4.2 shows some more detailed

specifications.

Figure 4.1: Camera and Lens connected together with Video

Cassette for size reference

CAMERA

Model: Jetcom Super Colour, High Sensitive CCD Camera

CCD: 1/2" Inteline-Tranfer Type

Lens Mount CS - Mount

Resolution 420 lines

Power Consumption: 4.5W

Dimensions: 51(W) x 43(H) x 128(L)

Weight: 380g

LENS:

Model: Computar H6Z0812

Focal Length: 8-48mm

Angle of view; vertical 5.22 - 33.11

Angle of view; horizontal 7.44 - 43.31

Minimum object distance 1.2mDimensions 57 x 95mm

Weight: 430g

Aperture F1 - F22Table 4.2: Detailed Specifications


15/52

The side controls of the camera are shown in figure 4.2, whilst the rear connectionsare displayed in figure 4.3.

The controls mounted on the camera are set in the factory for the best performance,

and thus should not need to be tampered with, but for completeness they are outlinedquickly here. On the side of the camera the shutter speed (ES), automatic gain control

(AGC) and white balance (WB) are situated. The settings should be kept on:

ES=6 AGC=ON WB=1

At the rear of the camera are the external connection points. Video output is a normal

75BNC connector, and the 12V supply is a low voltage socket. Also on the rear are

sockets for external sync inputs (when more than one camera is being displayed onone monitor) and automatic iris. These last two facilities are not used in this projectand are primarily designed for use in closed circuit TV situations.

The mounting of the lens is an important operation. If the lens is screwed all

the way into the lens mount on the camera, the expected range of focal points will beunobtainable. Instead, the lens should be screwed in only far enough to allow the

Figure 4.2: Side view of camera showing (L to R) Shutter Speed, AGC and

White Balance.

Figure 4.3: Rear of camera showing video output and DC voltage socket.


16/52

objects in the range of 1.2m - to be focussed on. This is an iterative process but

does not take more than a few minutes. Once the lens is in the correct place, it may befixed by tightening the hex screw situated on the base of the camera housing. The lens

and camera can be left permanently together since there is little reason to remove thelens from the camera when not in operation.


17/52

Film and video have certainly been used before, for decades in fact. Commonly thecamera (film or video) is static and views a length of a walkway, and records the

cycles of gait as the person walks from one extreme of view to the other. Often amarker is (or many markers) placed on the body so that the points may be recorded bycomputer (automatically or by an operator), which may then produce a wireframe

simulation of the locomotion, along which various statistics, as seen on the previouspage. Nowadays this is almost standard in any well equipped gait laboratory. Therehave been some apparatus which have had cameras mounted on trolleys and been

pushed by hand to follow and film a subject on the walkway, such as the examplebelow, at the University of Waterloo.

This project aims to combine these last two methods to produce a device

which will use a marker as a defining point to follow. That is, the device will aim tokeep the marker laterally in the middle of the video image at all times. Instead of justfollowing the person at walking pace and filming the lower body the instrument

conceptually will allow the study of a particular piece of anatomy. For example thecamera may film a close up of the knee (which therefore is where the marker would beplaced) in action, or the ankle, or a suspect prosthetic component in a fitted artificial

limb. The most important improvement that will be gained is increased resolution. Ifthe example of the kneeis continued; the knee would still be visible with a staticcamera, and with the dolly mounted version, but the cameras frame of reference

makes close and detailed examination of the joint difficult. Ideally with this newdevice the joint in question will appear almost laterally static on the video screen,allowing professionals to observe motion in a more convenient form. The objective of

the project is not to provide quantitativeinformation, but rather to improve the

clinicians subjective viewpoint of the patient. This qualitativeinformation cannot bemeasured to rate its success, but feedback from professionals should give anindication of the usefulness of this equipment.

Figure 1.2: Walkway and manually operated tracking cart, carrying both

film and video cameras. (Winter 1982, P409)


18/52

CHAPTER 5


To make the explanation of the circuit design more readable, it will bepresented in sections. The circuit diagram in each section will be shown and thenexplained, along with a description of the sections function and relation to the system

as a whole. The context of this chapter is such that it is assumed that someone may inthe future need to make changes or additions to the system, and thus explanation ofthe circuit is in sufficient detail to enable this without the party having to refer to a

great many external texts.

5.2 VIDEO SIGNAL EXTRACTION

This section forms the front end of the system. This is where the video signal

from the camera is fed in and the required information is stripped off. The circuitdiagram is shown in schematic #1. This section covers the the following material:

Format of video signals used

DC restoration Buffering of video signal

Colour burst stripping

Line sync pulse extraction Field sync pulse extraction

Intensing triggering

Contrast Triggering

Target latchingFirst it will be useful to outline the format in which the video signals are presented.

5.2.1 PAL VIDEO FORMAT

The upper waveform shown in figure 5.1 is a typical single line of pictureinformation from a PAL video source. (ie a camera, a VCR etc). PAL is the formatwhich is used in Australia, the UK and some parts of South East Asia to transmit

television and video information . (Other formats are similar but contain subtledifferences) The waveform shown was taken from the output of the CCD camera.

FIGURE 5.1: Example of a single line of a PAL video signal showing

Line sync pulses, colour burst and picture information.


19/52

5.4 TIMING AND DIGITAL TO ANALOGUE CONVERSION

This portion of the electronic system takes the signals created by the functionsdescribed in 5.1 and 5.2, and provides an output voltage proportional to the distance

of the marker from the left edge of the screen. An LED display is also included, whichtells the operator how many markers are currently being sensed; thus for optimaloperation only a 1 should be displayed. This section describes a circuit with the

following fucntions: Rouge image noise rejection

Region selection

Target latching Digital to Analogue conversion

LED display

IC10, an 74HC1648 bit shift register, has its reset pin connected to the

inverted field pulse pin (IC6, pin 4), its clock input connected to the line pulse output

(IC6 pin 5) and its serial data inputs both connected to the latched target pulse (IC9pin 3). This means that at the beginning of each field all of the shift registers outputs

are reset to logic low. At the end of each picture line, the value of the latched target (0or 1) is clocked into output Q1, and all other outputs are shifted along one output. The

eight outputs are connected to the inputs of a 74HC308 input nand gate (IC11). Theoutput of this gate will only go low when all of the outputs from the shift register are

high, in other words, a marker has been spotted on all of the previous 8 lines of theimage.

This is simply another precaution to exclude unwanted information from the

system. Single white pixels will therefore not upset the system since the next 8 lines

will not contain a marker, and the target information will be ignored. The output of the74HC30 goes to one input of a 74HC32OR gate (IC12). IC7:C is a NAND gate

which conveys information from the region selection (see section 5.2) circuits. Theoutput of IC12:D will only be LOW when the current line of the image is in the

selected region andthe option is selected, by setting the switch to logic high. Thisinformation is placed at the other input of the OR gate IC12:A. Thus when a target has

been found (and is more than 8 lines in height), the current line is in the scanningregion and the region select option is switched on, the output of the OR gate will belogic high.

IC9:B is a 74HC107JK flip flop, which is used to finally latch the target

signal once per frame. It is reset at the start of each frame by the inverted field pulse

(IC5 pin 4) and must be clocked at the exact instant that the marker is detected, notjust at the start of the relevant line. This is the reason why the output of IC12:A OR

gate is combined with the inverted pulse from the target latch (IC9 pin 2), so the pulsedrops at the correct time within the line of pixels. Figure 5.7 shows an illustrativeexample. The camera is looking at a table tennis ball.


20/52

5.7 AUXILIARY CIRCUITS

This final part of the chapter covers all portions of the circuit not coveredelsewhere. These are the 5MHz waveform generator, and LED position display.

5.7.1 CRYSTAL OSCILLATOR

This oscillator drives the 12 stage counter which provides the input to the 8 bit DAC.Thus the best level of resolution will be achieved if the 256 levels available are spread

evenly across the whole screen. If a time of 45S is conservatively set aside for a

whole line of pixel information, then the required clock rate for the counter is

fS

MHzc = =1

45

256

5 68

.

If a higher frequency is used there is the possibility that the counter will go over

binary 256, and the DAC will give a false output. The next available alternative is touse a 5MHz rate, which is achieved using a 10MHz crystal and dividing the result.The crystal oscillator circuit is shown in figure 5.13.

The output from IC7:B is thus a 10MHz square wave. A binary counter (74HC4040)is used as a divider to reduce the frequency to 5MHz.

5.7.2 LED POSITION DISPLAY

This module allows the operator to visually determine if the system is trackingthe marker as expected. A row of LEDs visually displays the horizontal position of

the spot which the circuit is triggering on, thus the

user can determine whether the circuit is detectingthe marker, or something else in the field of view.The unit is detectable from the rest of the circuit

board so it can be removed when not needed. The

circuit diagram is shown in figure 5.14. The moduleuses a LM3914Dot/Bargraph driver. This IC

Figure 5.13: Schematic diagram of 5MHz oscillator circuit.


21/52

CHAPTER 6

6.1 ULTRASONIC DISTANCE MEASUREMENT

This chapter outlines the work done on the ultrasonic distance measurementcircuits. Originally this was to be a part of the thesis assessment but as more emphasis

was placed on the video tracking portion, the time spent on this segment becamedetrimental to the more important aspects of the project, so research was stopped.However, a stage was reached such that when the rest of the project is completed at a

later date, some ultrasonics can be trailed with minimal further time investment.

6.2 BASIC CONCEPT

In air, at 20C, the speed of sound is 331.6ms-1. If a measurement of the time

in taken for a sound wave to travel from a point of emission to a receiver, then thedistance between those two points can be easily calculated. This is the basis for all

sonic distance measurement instruments. The speed of sound will alter will airtemperature, pressure and humidity, but the figure of 331.6ms-1is adequate for the

level of accuracy required here. Ultrasonicfrequencies are defined as those above the

audible range for humans, about 20kHz. Most ultrasonictransducers operate at 40kHz-50kHz. Rangefinders usingthese frequencies are commonly used in robotics,

industrial equipment, cameras and underwater vehicles.

All devices work by sending out ultrasonic waves from atransmitter, then waiting until they are reflected off an

object and returned to a receiver (which is most commonly

in the same location as the transmitter)

Most rangefinding equipment works in the following way: Transmitter sends a short burst of ultrasonic pulses

Digital timer starts

Receiver detects returning pulses and stops timer. Digital count processed further or displayed in distance units.

However, the purpose of the rangefinder in this project is to drive a motor so as tokeep the inline camera a constant distance from the subject being filmed on thewalkway. The motor driving circuits (which will use a Pulse Width Modulationmethod) will need an analogue input, so converting from analogue pulses, to a digital

timer, and then back to analogue output is quite wasteful. This is the reason that thisdesign instead uses a totally analog output method.

The new design approach is:

Transmitter sends a short burst of ultrasonic pulses Voltage ramp starts to increase linearly.

Receiver detects returning pulses and clamps voltage ramp to ground.

Peak detector maintains highest voltage level from ramp (with slow

discharge)

Ultrasonic transducers


22/52

CHAPTER 7

7.1 MOTOR SELECTION AND CONTROL

This chapter deals with the selection of the stepper motor and the circuits tocontrol it. A large amount of effort was put into designing a stepper motor driver fromdiscrete components, rather than simply purchasing a commercial unit. This chapter

outlines the designs that were built and tested, and why they were not chosen for thefinal design. Eventually a commercial unit was sought, and its operation is alsodescribed here.

7.2 MOTOR SELECTION

To determine the correct motor to use a calculation of the maximum expectedloads must be done. The maximum speed and acceleration can be obtained from gait

data.

Data will be taken from the ankle, since it experiences the highest velocitiesand accelerations. From experiments carried out at Rehab Tech, the time for one stepduring a brisk walk was found to be approximately 1 sec. Using the displacement

graph in figure 8.1, the gradient at the point of maximum slope is

vcm

ms=

= 100

1 432 32 1

sec %.

Calculations were done to check the suitability of several motors, but only thechosen model will be included here. The motor is a Size 23 Hybrid Stepper Motor,

which runs on a rated voltage of 5V, with phases rated at 1A. The specified maximumstep frequency is 880 steps/sec.

For a step angle of 1.8this amounts to =4.4rev/sec.

For a tangential velocity the effective radius of the gears will need to be

rv

m= =2

0 083

.

The specified torque for the motor is 500mNm, so the tangential force will be

FR

N= = 6 01.

Since the camera and lens weigh in total 810g (round conservatively up to 1kg), this

provides the moving camera with a maximum acceleration of 6ms-2.

Figure 7.1: Displacements of hip, knee and ankle over one step cycle [17]


23/52

CHAPTER 8

8.1 CONSTRUCTION

All of the circuits described were built onto a standard veroboard as commonly

used for prototypes. Since it is likely that some modifications will have to made inaccordance with later parts of this project, the circuit was not made into a dedicatedPCB. Room soon ran out on the first prototype board and was forced to use an IDC

connector between the two halves of the circuit. This undoubtedly has introducednoise into the circuit, which will not be present when the system is eventually put ontoone fabricated board. The LED position display, optical limit sensors, motor driver

and power supply are all separate from the main boards, and connect via plugs. Figure7.1 shows the first board.

Some features are highlighted:

Power in and indicators: Socket for 5V supply, and LEDs to indicate when

power is on.

BNC in:Composite video signal input socket.

Trigger Select:Switch to select between targeting on brightness or contrast

Trigger set:Potentiometer to adjust level of triggering

7 Segment display:Indicates how many spots the circuit is currently detecting.

Similarly for the second board; in figure 8.2 overleaf Pushbuttons: to select scanning region

Show targets:switch to select whether to display visual targeting information on

screen or not.

Scan select: switch to select whether to use scanning regions or leave then off.

Normal/info : display normal image or overlay scanning regions, targets etc.

To motor driver: plug which connects to motor driver board

Limit indicator:LED illuminated when optical limit sensors are triggered.

BNC out: composite video out.

Figure 8.1: First half of prototype board.


24/52

CHAPTER 9

CONCLUSION

The aim of this thesis project was to construct a system which would follow amarker placed on a subject, using a video camera, specifically for the implementation

in gait analysis. The system comprised of an electronic control circuit and amechanical trolley which would be driven by a stepper motor.

Unfortunately, motorised testing of the device was not able to be undertaken,

due to the unavailibility of the trolley (still being constructed). If the trolley was ableto be used then examples of actual tracking of an object, would have been able to bedone, rather than simply looking at a voltage output, or the motor turning whilst its

shaft is unconnected. However, great progress was made in the electronic circuitrydesign and construction, which was the main aim for thesis assessment. A list of themajor achievements is listed below:

Design of Ultrasonic rangefinding circuitry Construction and testing of Ultrasonic rangefinding circuitry

Testing and evaluation of commercial Ultrasonic rangefinding module.

Design and construction of Video Signal extraction circuitry

Design and construction of signal processing and display circuitry

Survey of available video cameras.

Survey of available zoom lenses for CCD cameras.

Design and construction of Motor control circuitry

Design and construction of two stepper motor driver circuits

Testing and Evaluation of commercial stepper motor driver.

Design and Construction of 15V, 6A overload protected power supply. Fabrication of all plugs and cables.

Design of Trolley mechanism.

The major part of the project, the design and construction of the circuits which :

Accepts video input from the CCD camera

Provides output for video display or recording

Has the user definable option of two triggering methods to track marker

Has a display showing how many markers are currently being detected.

Has a display showing the relative horizontal position of the market that thecircuit is detecting

Incorporates a method to reject any small patches of bright pixels on the image Makes the motor run at a constant velocity when more then one marker is

found, or no markers are found

Has user input and visual feedback on screen to define regions of the imagewhich are to be scanned, and those that are to be ignored.

Provide visual feedback as to bright areas on image which are being triggered

on.Rejects sudden jumps in marker position

Drives motor at correct speed and direction

Has facilities to stop trolley running off its track, yet does not deadlock the

systemAll of these features were designed by the author and work correctly.


25/52

BIBLIOGRAPHY

1. Braun, M (1992) Picturing time : the work of Etienne-Jules Marey (1830-1904).

Chicago : University of Chicago Press

2. Chelihi, N and Gaydecki, A (April 1993) A low-cost video-based multiple target

digital tracking system.Measurement Science and Technology, Vol 4, pp447-50.

3. Ciarcia, S (February 1986) Build an Audio and Video Multiplexer.Byte, Vol 11,pp85.

4. Ciarcia, S (July 1985) Living in a Sensible Environment.Byte, Vol 10, pp141-

158.

5. Dagognet, F (1992)Etienne-Jules Marey : a passion for the trace, New York :

Zone Books

6. Dolman, P (May 1989) Inside S-VHS.Electronics and Wireless World, Vol 95,

pp466.

7. Frerking, M (1978) Crystal Oscillator Design and Temperature Compensation,

Van Nostrand Reinhold Company.

8. Howard, P (February 1982) Camera Interface for a Microcomputer.Electronicsand Wireless World, pp37.

9. Keemink, G Snijders, C and Hoek van Dijke (January 1991) Upgrading of

efficiency in the tracking of body markers with video techniques. Medical

and Biological Engineering and Computing,Vol 29, pp70.

10. Kleeman, L (1995) A three dimensional localiser for autonomous robot

vehicles.Robotica, Vol 13, pp87-94.

11. Lenk, J (1988) Complete Guide to VHS Camcorder Troubleshooting and Repair.New Jersey: Prentice Hall.

12. Ma, C (September 1994) Direct Adaptive Rapid Tracking of Short ComplexTrajectories.Journal of Dynamic Systems, Measurement and Control, Vol 116,

pp537.

13. Milner ,M (1981) Human Locomotion: Data monitoring and processing-

clinical applications. (p123-155)Biomechanics of Medical Devices. New York:

Marcel Dekker Inc.

14. Newton, A and Maeder, H (Nov 1991) A TV analogue to digital conversion

system.IEEE Transactions on Consumer Electronics, Vol 37, pp829.


26/52

15. Omran, S (June 1985) Fast camera interface.Electronics and Wireless World,Vol 91, pp17-19.

16. Perkins, G and Clayton, K (August 1992) A low cost, general purpose S-video,

PAL/NTSC encoder.IEEE Transactions on Consumer Electronics, Vol 38,

pp512.

17. Rose, J and Gamble, J (Eds) (1994)Human Walking.London: Williams andWilkins.

18. Vaughan, C and Sussman, M (1993) Human Gait: From Clinical

Interpretation to Computer Simulation. (p53-66) Current Issues in

Biomechanics. Champaign: Human Kinetics Publishers.

19. Wezel ,R (1987) Video Handbook. Melbourne: Heineman Professional

Publications.

20. Winter, D (1982) Camera Speeds for normal and Pathological gait analysis.Medical and Biological Engineering and Computing, Vol 20, pp408.

21. Winter, D (1990)Biomechanics and Motor Control of Human Movement. NewYork: John Wiley and Sons, Inc.

22. Wobschall, D (1987) Circuit Design for Electronic Instrumentation. New York:McGraw-Hill.

23. (May 1988) Motors.Machine Design, Vol 60, pp57-97.


27/52

The only minor problem with the circuit is that of some noise in the DAC output.Over a 0-5V signal there is max 40mVpp noise, ie 0.8%. This problem is most

probably due to the wiring on the prototype boards, and thus the accuracy of thesystem will be even higher when the circuit is placed on a PCB.

In the current state of the project, as soon at the trolley is readied, the system

will be able to be tested, since the video, control, and motor driving portion all operateas designed. The larger design, involving the inline camera, motorised dolly andultrasonic distance measurement will be undertaken at a later date, but this area has a

solid foundation due to the work already done in this report on the ultrasonicrangefinding equipment.


28/52

The final version would also be housed in a box, which would have clearlylabelled controls and displays mounted. All components of the system would pug

together, thus providing easy storage when the system is not in use.

Shown are the two prototype boards, LED position display, optical limitsensors (on end of white cable), camera, and connecting mini coax cable. The motordriver board and power supply are not shown.

The design of the trolley to carry the CCD camera was completed early in the

year. The construction was to be done by the departmental workshop, butunfortunately the unit has not been completed at the time of writing, even though therequest was made in late July. The design for the trolley is shown overleaf in figure

8.4. The dimensions of smaller pieces are not shown as these were subject tocomponent availability in the workshop. The entire construction is to be aluminium tokeep mass low. The motor is coupled via gears to a driving wheel. A chain is looped

around this and is also connected to the actual trolley, such that it has a travel of 1.3m.This lenght was based on figure 7.1, in chapter 7. Notice the that peak to peakdisplacemeant of the ankle is about 80cm. An extra 50cm ws added to allow for

pathalogical gaits, unusual prosthetic devices and lag time in the dolly following thesubject.

Figure 8.2 Second half of circuit on prototype board.

Figure 8.3: Layout of system


29/52

The calculation of the gear ratio is as follows:Maximum velocity required, as calculated previously, is 2ms-1, but use 3ms-1

to leave space for any power loss in the device. The maximum speed of the rotor is

1000Hz at 1.8per step. This is (1000*1.8)/360 = 5 rev/sec.

In radians this is 2*5 = 31.4 rad/sec.

The maximum trolley speed is 3m/sec. The pulley is 15mm in diameter, thus therequired angular velocity is

=v/r = 3/.015 = 200 rad/s.

Therefore the gears need a ratio of 200/31.4 = 6.36 to obtain the maximum speed.

Motor, gears and pulleyCamera Trolley

1530mm

1300mm

The entire frame is to be made of aluminium forweight conservation. The transmission

methodwill be lightweight link chain, or plastictoothed wire, strung aroung both pulleys in acontinous loop and attached at the camera

trolley. At each end there shall be an opticallimit sensor (not shown) so that the device

cannot damage itself.The wheels will be nylon on bearings, and areangled such that sideways movement of the

camera trolley is eliminated.Below is an isometric diagram, showing

orientation of camera. (Lower diagram is not toscale)

Close up of Trolley, End view

100mm

150

mm

120

mm

80

mm

Nylon wheels with bearings

base

Camera secures here

30mm

Figure 8.4: Design of trolley mechanism


30/52


31/52

7.3 MOTOR DRIVING CIRCUITS

As mentioned in earlier chapters, stepper motors require that their coils beexcited in a specific sequence for motion to occur. Fortunately there exists integrated

circuits that take care of most of this bother. These chips require only the direction andstep inputs, and produce logic signals for the four phase outputs. For low powermotors, the coils can be connected directly to the IC, but since each phase in the motor

being used here can pass up to 1A, external components are required.Circuit #1 in schematic #6 shows such a circuit. The driving chip is

UCN5804B, which uses TTL level inputs. Eight MosFets are used to create fourinverting buffers, capable of switching the required voltage and current. The reverse

biased diodes are to protect the mosfets against voltage spikes due to inductiveswitching. This circuit worked, but the maximum stepping rate achieved was only300Hz, less than one third of the specified maximum of 1kHz. At higher frequencies

the voltage at the coils became chaotic and put the motor out of sequence, when

ideally the waveform at the coils should be square. Part of the problem was currentdrain, so more powerful power supplies were tried . This remedy improved

performance slightly but the maximum step frequency was still only around 400Hz.

The present mosfets (P channel IRF9520, N channelIRF510) were replaced

with new components with lower gate threshold voltages (P channel BUZ171, N

channel BUZ71) to make sure that the mosfets were switching properly. This did littleto aid the performance. On closer examination of the gate voltages, it was apparent

that the mosfets were indeed not being driven hard enough, even with the newcomponents. Thus circuit #2 (on schematic 6) was constructed. This used a new IC,

the SAA1027which runs on higher voltage levels, but still performs the same

functions as the UCN5804B. Thus the mosfets were driven much harder, (since theVgsis now much higher), and performance increased accordingly. The maximum ratewas now 700Hz. Continuing on this avenue of research the lower supply rails of the

SAA1027 were made negative, so that the P channel mosfet was also driven harderthan before. This also improved performance, but the rotation of the shaft was notsmooth at high speeds and the available torque was quite low.

After some more experimental circuits, some deeper research into higherperformance stepper motors was done. It was found that for better performance acurrent regulated circuit was needed. That is, instead of simply switching the coils to

the 5V rails, the circuit also needs to control the current through each coil, keeping itconstant. One elementary way to perform this function was trialed, which is to use

higher voltage with a resistor to drop the voltage down to the motors specifications,but without added control circuitry this made little difference. This technique is calledthe Unipolar L/nr method, and is described further below. Fluctuation of phase currentwas the primary cause of the non-performance of the previous circuits. A consultation

with Mr Holmes of the power electronics group was made, and he stated that simplywith a lack of power background the construction of such a circuit would beimpossible. Indeed, it could nearly be a thesis project in itself. Thus attention was

promptly turned to the option of commercial motor drivers, which incorporate currentregulation, and obtained a RS Unipolar Stepper Motor Driver Board.

The board connects using a standard 32 way 41612 socket (as used with

Eurocards). The module uses a driving method called Unipolar L/nr drive, and is aone way to obtain some standard of current and voltage regulation. nR is the sum of


32/52

external resistance plus winding resistance (R), where n is a real number. By selectinga higher value for n (ie: larger external resistance) and using a higher DC supply to

maintain specified voltage and current on the windings, the torque and speed ratingsare improved. Coupled with other control circuitry this module provides excellentperformance with speed up to 1200Hz (over the motor specifications) and high torque.

For the motor purchased, (5V, 1A) at a supply voltage of 15V, therecommended resistor value is 10, rated at 10W.

Since the control signals from the system are at TTL levels, whereas the motordriver runs at 15V, an open collector TTL Nand gate is used to interface the two. This

is the 74HC03. The connections to the driver module are shown in schematic 7.

Also required for the driver is a 15V power supply. A suitable supply wasconstructed which is capable of up to 6A current, and is overload protected. Thecircuit for this is shown in schematic #7.

The voltage is dropped from 240VAC to 17.5VDC and smoothed, rectified. A

three terminal 15V voltage regulator is used (LM7815) to provide the required output.These regulators are rated to 1A, which is satisfactory according to the motorspecifications. However, in practical situations stepper motors can require much

higher peak current supply, over 4A in a worst case. Such instantaneous current spikeswould cause interruptions in the supply due to the internal overload protection in theLM3815. Therefore a current boosting transistor is used to provide greater current

availability. This is P3, a TIP2955. When current through the regulator is over 90mA,the VBEexceeds 0.6V and the transistor switches on, carrying all excess current.

Furthermore, to ensure that any accidental short circuits do not damage the

supply or motor driving circuitry, a current limiting feature has been included.Resistor R5 sensed the load current being delivered, and when the voltage drop acrossit exceeds about 0.6V P2, another TIP2955, cuts off the drive current to P3, thus

initiating the automatic cutoff in LM7815. For 0.6V to appear over R5, a current of6A must be sunk. This is enough to cover all requirements for the motor, but will notdamage the supply circuit or driver.

Figure 7.2: View of commercial stepper motor driver board


33/52

The peak of the voltage ramp will be in proportion to the time taken to send andreceive the pulses, thus if pulses are sent and received regularly the output of the peak

detector will follow proportionately the distance of the object from thetransmitter/receiver pair. The block diagram of the basic circuit is shown below infigure 6.1.


The circuit diagram for this first prototype is shown in schematic #5.The base ultrasonic frequency is produced by a crystal oscillator and a divider. The

crystal oscillates at 5.0688MHz, using a biasing circuit similar to that described in

chapter 5. A 404012 Stage Binary Counter is used to divide this clock input by 128(by connecting an output to Q7) which gives an output square wave frequency of39.6kHz. This is acceptable since the receiver and transmitter being used operate in

the range 401kHz.

The NE555timer is configured to give an output pulse at a frequency of4.34Hz, with each pulse going logic low for 0.69mS. This pulse and the 39.6kHz

frequency are combined using a 4011Nand gate. Thus the output of this gate is a

0.69mS burst of 39.6kHz signal. This is enough time to transmit 27 pulses, which iseasily enough for the receiver trigger on. The frequency with which the pulses are sentinduces a limit on the maximum distance that the device can measure. Pulse bursts are

sent at 4.34Hz, which is every 0.23sec. The maximum distance that can be covered inthis time is

d ms s m= =331 6 0 23 76 41. . .

Since the sound must be reflected back, the actual measurable distance is 35.9m, butthis is quite ample considering the low power of the transducers restricts their

maximum range to about 10m (5m reflected). Spare time after the reception of thepulse is advantageous since rogue reflections (off distance walls, roof etc.) in alaboratory environment can sporadically interfere with the reception of the correctpulses, so these can be dissipated in the extra time before the next pulse is transmitted.

The ultrasonic transmitter is driven by a 2N222transistor.

The signal from the ultrasonic receiver is amplified by two 741 op amps.Figure 6.2 shows various waveforms taken from the circuit. The top trace is the pulseoutput of the NE555, and the next trace is the output of the Nand gate, showing the

short burst of 39.6kHz cycles that are produced. Trace three is the beginning of theamplified receiver signal, created by the reflection off an object about 20cm away.

40khzgenerator

pulsetimer

ramp

receiver

low passfilter

voltage outputproportionalto distance

Figure 6.1: Block diagram of basic Ultrasonic rangefinding circuit


34/52

From the trace, the delay from the transmission of the pulse, to its arrival at the

receiver is 1.2mS. Thus the distance of the object from the transmitter/receiver is

d cm= =331 6 1 2 101

219 83. . . .

In the first version of the circuit a phase locked loop was used to trigger when thepulse appears at the receiver. This PLL was in the commonly used NE567 tone

decoder, which is capable of a large range of locking frequencies, and has adjustabledetection bandwidth. This component can be damaged by supply voltages over 10V,

so a LM3808was used to create a secondary supply of 8V to the PLL and the two

preceding amplifiers. The rest of the circuit runs on 12V to obtain the maximumpower output from the transmitter, which has a rating of 12V. Components C10 andP1 are used to set an internal reference frequency that the NE567 uses to compare to

the incoming signal. This internal frequency is governed by

fR C

o

X X

1 1.

BWV

f C

I

o X

= 1070 (% of f o)

BW is the detection bandwidth. VIis the amplitude of the input signal, 2.5V in thiscircuit. In this case fois set to 40kHz, and the bandwidth is 9.3%. The NE567 takes 8-12 cycles to lock onto an incoming signal. The circuit performed with the tone

decoder, but the output of the circuit was spasmodic, even when the experimentalconditions were static. Unfortunately the PLL was not congenial toward the shiftingamplitude present in the receiver signal, even if the frequency was constant.

Adjustments were made to the receiver amplifier but this did little to improve theperformance.

At this the necessity of the PLL in the receiver was questioned. The ultrasonic

receiving transducer only resonates in the 401kHz range, so the problem of audio

interference will not be a problem, and the tone decoder is not needed to trigger on theamplified pulse when another device could do the same task by triggering on the

voltage levels, and do away with the tuning circuitry need for the NE567. Thus MKII

of the circuit used a comparator to produce a suitable output for the receiver. TheLM311 (IC4) was used, and the output of this chip is shown as the bottom trace in

Figure 6.2: Traces from ultrasonic rangefinding circuit. From top: Timing pulse from

NE555, input to transmitter, received signal after amplification, and signal aftercomparator.


35/52

figure 6.2. The results from this new design were much more encouraging, so now theanalogue timing portion of the circuit could now be developed.

For the ramp a simple RC circuit was used, with an analogue switch used todischarge the capacitor and reset the circuit. A very long time constant is used, so thatonly the first few percent of the rise time are ever used as a ramp, and this part of the

exponential is linear enough to model a voltage ramp. The RC combination is madeup of R16 and C11, and using a 4016 Quad Bilateral Switch for the discharge. The4016 switch is controlled by a 4013 Quad D type flip flop. This uses the starting pulse

of the NE555, and the finishing pulse of the comparator to provide a logic low at itsoutput from the time of transmission of the ultrasonic pulses to their arrival back atthe receiver. This turns the analogue switch off for this period, allowing the ramp to

rise (almost) linearly during this time. Figure 6.3 shows this operation.

The top trace is the pulse provided by the NE555. Channel 2 shows thereceived pulses after being amplified and passed through the comparator. (Note that

the pulses are not clear on this channel due to aliasing in the digital oscilloscope atthis relatively low scanning rate.) The forth trace shows the output of the D flip flop,thus controlling the period of the ramp, shown on channel three. It is obvious that the

longer the delay between transmission and reception, the longer the ramp time and thehigher its peak voltage, which will increase by 0.65V/mS (adjustable by R16). This

ramp output is buffered by the voltage follower created by a 741 op amp. Finally theoutput is passed into a crude peak detector with slow discharge via pot R17. Thisproduces a sawtooth waveform, the average of which is proportional to the distance ofthe closest object from the transducers.

This final output could be passed through a low pass filter for smootherresults, but this was not done on this initial prototype. This circuit provided betterresults than the design using the PLL, since the comparator always triggered on the

returned signals. However, the range of the device was still no more than 1.0m for areliable signal output. Further research indicated that these type of transducersinherently have a high Q, and performance can vary greatly over a transmission

frequency range of a few hundred hertz, a fact not mentioned in the specifications. A

variable frequency generator unit was used to replace the initial 39.6kHz generator, sothat performance could be tuned with frequencies around 40kHz. Aswell, new and

Figure 6.3: Traces from the ultrasonic Circuit. from top; start pulse from NE555, outputfrom receiving comparator, voltage ramp, output from flip flop.


36/52

higher quality transducers were obtained. This resulted in some extra sensitivity (up to2m), but the reliability was still insufficient for the device to be incorporated into the

rest of the project. It is also important to note that the laboratory experiments weredone by reflecting the sonic signals off hard objects such as books and walls. Humanbeings unfortunately are not quite so substantial, especially in a clinical situation

where subjects are more likely to be donning loose fitting attire. It was at this stagethat a decision was made to discontinue the development of this circuit in favour ofmore pressing parts of the project. However a few ideas for further development were

entertained, but time will not permit experimental validation of these concepts. Thesewere:

Use a solid reflector (eg: stiff plastic) on the subject to aid reflections.

Separate the transmitter and receiver. The transmitter could be batterypowered and clipped onto the walking subject. At the same instant that theultrasonic pulses are transmitted, an ultraviolet signal could also be

transmitted by an array of UV LEDs. The receiver would time from the

reception of the UV pulse to the arrival of the ultrasonic signals. Thiswould totally omit the need for a solid reflecting surface, but the circuitry

and battery power will make for a more complicated circuit. Try other ultrasonic transducers available.

This last notion was briefly explored when, along with another student, an

investigation of a commercial sonar ranging module was carried out. The exact name

is the Polaroid 6500series, and the device contains the a transducer (transmitter andreceiver combined) and driving circuitry. The module runs off a 5V supply, andaccepts a driving pulse from the user, at which point it will transmit pulses, process

the returned signal, and provide a logic output high when the pulses have returnedsatisfactorily. Unfortunately the unit requires high current for the period when

transmission occurs, of over 2A. For normal power supplies this is too great a drain,and as a result the supply voltage drops instantaneously, resetting the whole circuit.

Thus the output is never triggered. Experiments were conducted with morepower supplies and smoothing capacitors on the supply rails, but the unit still failed to

give correct results. Eventually an idea was trialed which was to bypass a section ofthe modules circuitry and take a signal (which pulses high when the module senses thereturn waveforms) directly from a pin on an internal IC. This worked well and is easy

to implement, requiring only one external logic IC. The module (costing around $50)offers a high range (we tested to over 7 metres) and good reliability, so this is a goodpossibility for the remainder of the project to be completed at a later date.


37/52

simply transforms an analogue signal into a linear display. Pot P3 is used to adjust thevoltage level which corresponds to full scale deflection.

Figure 5.14: Schematic of position indicator circuit


38/52

Note that this oscilloscope trace covers one entire field. The top trace is the outputfrom 8 input nand gate IC11, showing the set of lines for which the marker is visible.The second trace is the output from IC7:D, showing the region selected for scanning

by the user. This has been purposely place so that half of the ball is in the region,whilst the other half is outside, for the purpose of illustration. Channel 3 is connectedto the output of OR gate IC12:B, displaying the combination of the latched target

pulse, but onlywhere the object is within the selected scanning region. Finally, the

bottom trace shows the latched JK flip flop, which remains latched high until thebeginning of the next field. The bottom two traces bear closer examination. Figure 5.8shows more detail of the same example.

The top trace shows normal PAL picture information. The location of the ball isquite obvious here. The second trace is the output of OR gate IC12:A; after 8 lineswith target latches have been put into the shift register, the output of the OR gate

finally goes low (if the lines are within the selected scanning region). The third traceis from the inverted output of the OR gate IC12:B. This combines the latched targetpulses, and thus the JK flip flop latches at the instant the target (the ball) is

Figure 5.7: Trace showing (from top) pulse after shift register, scanning region,

combination with latched target, and final latching for whole field.

Figure 5.8: Trace showing picture information lines, output of OR gate, combination

with latched target and latched JK flip flop.


39/52

encountered. This is vital since the output of the JK latch will be controlling the inputsto the DAC.

This latching time is so important because the noninverted output of the flipflop (pin 5) latches binary time information from a 12 stage counter. The counter used

is the 74HC4040(IC13), and the latch is a 74HC374octal D type flip flop (IC14).

The counter is driven by a 5MHz crystal oscillator, and is reset at the beginning ofeach picture line. The first eight binary digits are fed directly to the octal latch(controlled by the JK flip flop as above), and the outputs from the latch continue to theinputs of the digital to analogue converter.

This component is the commonly used DAC0800. From above it can be seenthat binary information is latched through once in every field, when a target is found.If no target is detected the latch remains in its previous state, and thus the DAC output

remains at the same level until a target is located. The DAC0800 has dual currentoutputs, so a LM741 is used to convert this to a voltage waveform, producing the finalvoltage output at pin 6. An example of the output is shown in figure 5.9. The camera

is observing the motion of a table tennis ball swinging in front of the camera over aperiod of 50 seconds. The sinusoidal motion and the exponential decay of theamplitude are quite visible. The fist few periods of motion appear to have equalamplitudes; this is because the ball swings momentarily out of the image, thus the

signal appears clipped.

As mentioned at the beginning of this section 5.3, the circuit includes a 7-

segment LED display to indicate how many markers the system can detect in thecurrent image. The number found is kept in a binary counter, another 74HC4040

(IC17) which is reset at the end of every field. IC18 is a 4511BCD to 7-segmentlatch/driver. The major problem in this part of the circuit is that the latch enable (LE)on the 4511 must be driven logic low (to allow new data into the latch) then high (to

lock in the data) beforethe counter is reset. Since all of the available sync pulsesoccur simultaneously, some new pulses had to be created. This is where the extendedfield pulse is used. This longer field pulse is logic ORed with the short field pulse to

provide a negative going pulse (used to latch the 4511) which is just after the normalfield pulse (used to reset the counter). Figure 5.10 illustrates:

Figure 5.9; Example output of DAC, showing exponential damping of a pendulum.


40/52

The top two traces are the inputs to the OR gate. Trace 3 is the inverted short fieldpulse, used to latch the data into the display driver. Trace 4 is the output from the ORgate, and resets the binary counter just after the latch has occurred.

5.5 MOTOR PULSE GENERATION AND SAFETY

This section of the system takes the signal from the DAC and transforms itinto outputs that the stepper motor driver can use. Stepper motors have a number of

coils (usually four) which are each externally accessible. The internal mechanics of

the motor are such that the phases must be excited in a certain sequence for the shaftto rotate. The advantage of these motors is that each step in the sequence rotates the

shaft by a specified amount, thus they are extremely useful for precision work sincethe position of the end effector is known to some accuracy, but complicated shaftencoders and feedback loops are not required for good results. They are also popularsince their design is very compatible with digital control. The motor used in this

project has a rotation of 1.2per step.

All stepper motor drivers require at least two inputs to control the motor, step

and direction. Often a half/full step option is also offered, where the half stepsequence is more complicated but offers better resolution that the sequence of phase

excitation used in the full step control. Naturally the faster the steps, the faster themotor rotates. Thus the DAC 0-5V signal output must be transformed into a directionand step output for the motor driver. The dynamic control system will initially be a

simple proportional output; ie the further the marker is from the middle of the image,the faster the motor will turn in the opposite direction to correct this. This willcertainly have to be improved to a more complex control system, but without the

trolley on which to mount the and test camera on, this is a futile task. The signalconditioning is represented in the block diagram below;

Figure 5.10; From top: inverted long field pulse, short field pulse, inverted short

field pulse, result of OR between traces 1 and 2.


41/52

The signal is first put through a low pass filer. This is to reduce the impact of

any instantaneous disruptions in the signal, from momentary reflections, the markerbeing obscured, and so forth. The level has filtering has been carefully selectedbecause if the -3db point is too low, the RC circuit will introduce a phase difference

that will effect the dynamic response of the whole system. An offset is then introducedto the signal. The DAC output is 0-5V, but for upcoming parts of the circuit a signal isrequired that is centred about zero volts, so the offset (which is adjustable) shifts the

waveform while degrading it by a factor of two, to prevent clipping. A comparator isthen used to extract the direction signal. The offset signal is then rectified (to obtain aspeed magnitude) and finally passed to a VCO which produces pulses with a

frequency proportional the voltage magnitude. The circuit is shown in more detail inschematic #4.

The lowpass filter is made from R18 and C4. The offset and degradation are

achieved simply using a LM741 (IC32). The zero volts point is now the reference

point for the centre of the image, since the speed and direction of the motor aredetermined by the amplitude of the signal voltage with respect to this point. Thus thevariable offset is useful, for the operator can define where on the image s/he wishes

the system to centre the marker on. The rectification is slightly more complicated. Asimple diode rectifier was insufficient since the 0.6V voltage drop inherent with suchdevices will distort the waveform. Thus active rectifiers were used, made up by

IC33A and B, two halves of a LF412dual fet op amp. This signal is then amplified byIC35, another LM741.

The direction in which the motor must turn is determined by the comparator

IC34, a LM311 comparator. Simply, if the signal is negative the motor turns in onedirection (at a speed determined by the amplitude), and if it is positive it turns in theother. The comparator is fitted with a 150mV hysteresis, because when the signal is

around zero volts, any noise would cause the motor to change directioninstantaneously, interrupting the tracking. The hysteresis is induced by R13 and R14.

The hysteresis is controlled by the equation

L+and L-are the saturation voltages of the comparator (+5V and -5V in this case).Values of R13 and R14 were chosen to be 47K and 1.5K respectively to give a

LP FilterDACsignal

Reductionand Offset

ActiveRectification

VCO

Comparatorwith

HysteresisDirection

Speed

Figure 5.11: Block diagram of signal conditioning circuit

L-

L+

V t2Vt1

V LR

RT1

13

14

= +

V LR

RT2

13

14

=


42/52

hysteresis of 150mV. Since the rails of the comparator are 5v, the output from pin

7 will also swing between these two rails. Thus transistor N1, a NPN548 is used therestore the output to 0-5V levels.

Finally the amplified signal from IC35 is placed at the input of IC36, a LM331Precision Voltage to Frequency Converter. The output of this component is governed

by

fV R

R R Co

IN=

9

11 7 12 09.

where Vin is the input voltage at pin 7. In the configuration shown in the schematic,

the range of frequencies available when the input is 0-5V is 1-1200Hz. Over this

range the linearity of the VCO is 0.01% of full scale.

Figure 5.12 shows an example of this circuit output. The uppermost waveform is theoutput of the DAC, created again by a pendulum infrom of the camera. Trace twoshows the rectified waveform (not here that the initial signal has been offset so that

the zero volt point is at the midpoint of the sinusoid.) The bottom trace shows theoutput of the comparator, showing the direction changes.

5.6 LIMIT SAFETY FOR TROLLEY

It is vital to provide some protection on the trolley mechanism for occasions

when the motor attempts to drive the trolley further than physically possible on thetrack. This is acheived using two UV optical slot sensors, one at each end of thetrolleys track. Rectangles of opaque material shall be placed on the moving trolley, sothat when it is too close to either end, the material will enter the slot and interrupt the

light beam. These optical devices are available already in suitable packages. Theelectronics to do this is also shown on schematic #4. Each sensor has its emitter driven

by 20mA, and the collector of the output transistor drives the base of a PNP548transistor, so that an interruption of the beam causes a positive output from the

collector.The two signals from the sensors are fed into an array of logic gates, along

with the direction signal, and the speed output. The efftect of these gates is that if thetrolley is travelling in one direction and exceeds its limit the speed output is disabled

Figure 5.12: Trace of DAC output, rectified signal, and direction output

from comparator.


43/52

until the direction input changes, at which time the speed output is once againenabled. Thus if the trolley travels too far along the track, it simply stops and waits

until the dirction signal instructs it to go in the opposite direction, instead ofsimplistically stopping and deadlocking the entire system.


44/52

A video picture is made up of 625 lines, but to save bandwidth they are

transmitted alternately, in two sets called fields. Every second line is transmitted inone field, then the remaining lines are transmitted in the next. (These are also called

the oddand evenfields) Since the interleaving lines are still transmitted at 50Hz, anddue to the persistence of the phosphor screen, our eyes cannot perceive the difference.

Thus each field contains 312.5 lines of pixel information. To enable the televisionmonitor to correctly place the picture, the source generates various pulses tosynchronise the electron guns in the picture tube.

Before each line of picture information there is a line synchronisation pulsewhichinstructs the electron gun to turn off and sweep to the beginning of the next line of

pixels. Also present, just after the sync pulse (in a region called the front porch) is a

short sequence of pulses called the colour sync burstor chroma burst. This series of8-10 cycles is used in colour televisions to help decode the chrominance (colour)component of the signal. It is actually at a very high frequency (many MHz) and so is

not clear on this trace.

The colour component of the picture signal (which is transmitted at the sametime as the intensity information) is phase modulated at a frequency over 4MHz, and

the colour burst is needed to provide a reference phase at the beginning of each line.

After the colour sync burst is the picture information, which takes up about 40S of

the 65S period. The magnitude of the picture waveform determines the intensity of

the pixels at that point along the line, ie. a higher voltage produces a bright pixel.There is also colour information in this section of the signal but again it is at very highfrequencies, so is not visible in the diagram.

After the 312.5 lines have been displayed on the screen, the video generatortells the monitor to return to the top of the screen by transmitting a series of short

pulses. This is shown in figure 5.2. (Note on the left side there is half a line of pictureinformation at the end of the last field. The next half is at the beginning of the next

field) This section of the signal is called the field synchronising pulses. After thisgroup of pulses has been sent, the next 312.5 lines of picture information in the next

field are transmitted.Now the scheme is continued to extract this information in a useful way. It is

required to somehow distinguish the anatomical marker from the rest of the picture,and there are two way that have been implemented to do this.

FIGURE 5.2: Trace of PAL signal showing lines of Picture

information and start of the Field Sync Pulses


45/52

The first is to assume that the marker is the brightest part of the picture. Sincethere is a signal which has a voltage level directly proportional to luminance

(picture intensity) a voltage comparator can be used to pick up the brigh

Video Tracking fot Gait Analysis

Documents