measurements and signal processing (part 2) mce 493/593 & ece 492/592 prosthesis design and...

22
Measurements and Signal processing (part 2) MCE 493/593 & ECE 492/592 Prosthesis Design and Control September 30, 2014 Antonie J. (Ton) van den Bogert Mechanical Engineering Cleveland State University 1

Upload: jaylin-motte

Post on 14-Dec-2015

218 views

Category:

Documents


4 download

TRANSCRIPT

1

Measurements and Signal processing (part 2)

MCE 493/593 & ECE 492/592 Prosthesis Design and Control

September 30, 2014

Antonie J. (Ton) van den BogertMechanical Engineering

Cleveland State University

Today

• Laboratory techniques for human motion– Camera-based motion capture– Force plates & instrumented treadmills– Balance testing– Strength testing

• Lab tour– 7:20 PM– FH 269

History of motion capture

• Muybridge, 1870s– multiple cameras, 2D

• Marey, 1870s– strobe lights as markers

• Braune & Fischer, 1895– strobe lights, 3D

3

4

Distance-based measurement

• Measure distance to three (or more) sources– solve XYZ from 3 nonlinear equations with 3 unknowns

• GPS– resolution insufficient for human motion

• Ultrasound– www.zebris.de

5

Active marker systems

• Markers are LEDs– flashing sequentially

• Camera– projects marker on image

plane or line

• Most common: three 1-D cameras in one box– high resolution– high frame rate– markers must be seen from

box

Optotrak

Codamotion (no lenses!)

6

Passive marker systems

• All markers visible– 2D cameras

• 16 mm film, analog video– manually digitized

• Digital video cameras– reflective markers– infrared strobe lights– high contrast, thresholding– 2D marker centroid coordinates

• combined into XYZ of markers

– Vicon, Motion Analysis, Qualisys

3D measurement requires at least two (2D) cameras

lensx

z

y

image plane

3-D space

vu

camera model:DLT (direct linear transformation)

1

1

11109

8765

11109

4321

zayaxa

azayaxav

zayaxa

azayaxau

a1…a11 are calibration constants(different for each camera)

Two cameras: u,v are measured in each camera

Solve x,y,z from 4 equations

More cameras:• better accuracy• less chance of marker loss

Capture Labat Electronic Arts:132 Vicon cameras

Fenn Hall 269:10 Motion Analysis cameras

9

Recent developments

• Markerless motion capture• Improved IMU data processing• IMU combined with range sensor – www.xsens.com

• Microsoft Kinect

• Optical, camera-based measurement with markers is still the “gold standard” for human motion labs– still very expensive

vu

Camera-based motion capture in 2D

lens

camera image planeparallel to XY plane

markersassumed tostayin XY plane

y

x

Camera model:

O

O

vv

uuS

y

x

cossin

sincos

Camera parameters:S: scale factor (meters per pixel)θ: angle between X-axis and U-axisuO,vO: image coordinates of XY origin

determined by imaging a rodof known length, one end at origin,aligned with X-axis

Matlab code for measuring U,V from videomovie = VideoReader(‘testfile.avi'); % load the video filenframes = movie.NumberOfFrames;height = movie.Height;npoints = 10; % how many points must be measured in each frameuvdata = []; % make a matrix to store the data

% display each frame and measure U and V of all pointsfor i = 1:nframes

d = read(movie,i); % extract frame i from the movieimage(d); % put the image on the

screendisp(['Frame ',num2str(i),':']);disp(['Click on ',num2str(npoints),' points']);disp('Click to the left of the image to stop.')g = ginput(npoints); % collect data until user has clicked

on all pointsif (min(g(:,1)) < 0) % if any point had a negative U-

coordinate, stop breakenddisp('Done')g(:,2) = height - g(:,2); % invert V coordinates so V-axis will point

upwarduvdata = [uvdata ; reshape(g’, 1, 2*npoints)]; % add a row to the data

matrixend

Clinical Orthopaedics andRelated Research, 1983

Techniques used:• 16 mm film at 50 frames per second• camera car alongside walking subject• markers on wall behind subject for calibration• Numonics Digitizer & microcomputer• IBM 370 for processing• about 2 mm random error in coordinates• 5 Hz low pass filter

Angle measurement

Two markers on a body segment segment angle Joint angle = difference between two segment angles

Winter, 3rd Edition, Fig. 2.31

Matlab:theta21 = atan2(y1-y2, x1-x2);theta43 = atan2(y3-y4, x3-x4);theta_knee = theta21 – theta43;

• atan would give results between –π/2 and π/2, requires extra “if-then” logic

• atan2 function gives results between –π and π, can represent full range of rotation

• use “unwrap” function on time series if angle jumps between –π and π

If you use Excel:

43

4334tan

xx

yy

Some real data

What is the knee angle at time = 2959.594329?

1: RGTROright greater trochanter

2,3: RLEKright lateral epicondyleof the knee

4: RLMright lateral malleolus

theta21 = atan2(y1-y2, x1-x2);theta43 = atan2(y3-y4, x3-x4);theta_knee = theta21 – theta43;

theta21 = atan2(0.90533-0.51603, -0.19465-0.01730)theta43 = atan2(0.51603-0.12862, 0.01730--0.09302)theta_knee = theta21 - theta43

X

Y

Force plate

• Measures ground reaction forces– rigid plate supported by four (or three) 3D force sensors– main vendors: Kistler, AMTI, Bertec– measures 6 variables: resultant 3D force (Fx,Fy,Fz) and moment

(Mx,My,Mz) on the axes of the force plate– also available as instrumented treadmill– http://www.kwon3d.com/theory/grf.html

AMTI

(a) Fxyz, Mxyz(b) forces acting on foot(c) forces in load cells(d) force and torque acting

at center of pressure (COP)

Equivalent force systems:(b) = (c) = (d)

Fx,Mx Fy,My

Fz,Mz

Resultant 3D force and moment from four load cells• 3D force F, applied at r, is equivalent to a 3D force F

applied at the origin, plus a 3D moment M = r x F• Resultant of all four:

4321

4321

MMMM

FFFF

M

F

COP (center of pressure) representation

• 3D force F is assumed at COP rather than origin• Definition of COP (x,y)– z=0 and Mx=My=0 at COP (zero moment point)

• Remaining moment Tz about vertical axis– “free moment”

• still 6 variables

xyzz

z

x

z

y

yFxFMT

F

My

F

Mx

DIY GRF measurement(and save $50,000)

Brodt et al. (2013) Instrumented foot bar for Pilates exerciseXXIV ISB Congress, Natal, Brazil

Simple force plate

• Vertical force only• Three points of support (no static indeterminacy)• Gives accurate COP in certain conditions

(Zsensor * Fx << My and Zsensor * Fy << Mx)

FORCE

Zsensor

Instrumented treadmills

• Treadmill frame sits on three or four 3-axis load cells– must be stiff and light

• Separate belts for left and right• Very good for clinical research

– each step is a measurement– speed can be controlled or self-paced– weight support is possible

• Prosthetics research– controlled speed– prosthetic device can be tethered to power supply and computer

ADAL treadmill at Cleveland VAMedical Center

Strength testingMaximal isometric torque

force fromleg

motor andtorque sensor

Isometric test: constant joint angleIsokinetic test: constant joint angular velocity

Speed dependent torque

muscle shortening(concentric)

lengthening(eccentric)

Cybex

Kincom

Balance testing (clinical)Platform with controlled rotationBuilt-in force plate (vertical force only?)COP calculation

• screening for risk of falling• balance training• knee injuries• concussion testing

Biodex SD $12,500 http://youtu.be/cBBlTYMulsE