Block Diagram of Testing Jig

1
A Body Sensor Network for Tracking and Monitoring
of Functional Arm Motion
Kim Doang Nguyen I-Ming Chen Zhiqiang Luo
Song Huat Yeo Henry Duh School of
Mechanical and Aerospace Engineering Nanyang
Technology University, Singapore 2009 IEEE/RSJ
Intl Conf on Intelligent Robots and Systems
(IROS) St Louis, MO, USA October 12-14, 2009
2
Motivation

• Objective measurement of motor recovery in
  patients
• Assessment evaluation of patients strength,
  range of motion, muscular activation patterns in
  recovery
• (Subjective) Measures impairment (Fugl-Meyer
  assessment), daily activity (Barthel index)

• Compact, user-friendly measurement devices
• Accurate, easy to wear, ergonomic, hygienic,
  inexpensive
• For clinical and personal uses
• Motivating patients for rehab program/sessions
  with suitable interactive applications

3
Outlines

• Motivations and background
• OLE sensing module
• Body sensor network for OLE SmartSuit
• Sensor placement and arm kinematic model
• Experimental results and statistical tests
• Conclusion

4
Existing Motion Capturing Technology
Outside-in
Inside-in
5
Optical Linear Encoder Sensor Joint Flexion
Angles

• Principle of joint flexion angle measurement
  sensor
• Opto-mechanical strip encoder in bend guard (soft
  exo-skeleton)
• Single axis linear motion (synthesis for
  multi-dof)

300 lpi encoder 0.1 deg res. (avg R 50mm)
6
Prototype of OLE Sensor

• Encoder AEDR-8400-132 (Avago)
• Accelerometer LIS3LV02DQ (STMicroelectronics)
  (640Hz)
• Microcontroller dsPIC33FJ32MC204 (Microchip)
• CAN controller MCP2515 (Microchip)

Operating voltage 3.3V Sampling frequency
15kHz Accuracy 0.1flexion angle Operating
current 7 mA Maximum reader speed 1.5 m/s
7
Optical Linear Encoder SmartSuit
8
Body Sensor Network
SmartSuit sensor network for capturing arm
motion three sensor nodes and one data
concentrator, connected via CAN bus
RF station
9
Human Arm Kinematic Model
3 OLE accelerometer to obtain 7-DOF
Arm Shoulder (3) Elbow (1) Wrist (2)
1

• At shoulder, ?0?w0. ?w1 from Acc-1

• ?1 calculated coord transf

• At elbow, ?2 from OLE-1?3 from OLE-2

2

• Expand trans matrix from world to frame 4 gives?4

• At wrist,?5 from OLE-3?w4 ?w6 from Acc-3

3

• Expand trans matrix from world to frame 6 gives
  ?6

10
Smart Suit App Demo
11
Validation of OLE Sensing Modules
To testify working principle of OLE
To appraise performance with rigid links and
joints, known joint diameter and center.
Jig diameter 63 mm PowerCube
0.180/pulse Encoder 2000 counts/rev
12
Validation of OLE Sensing Modules

• Three sets of measurements

• Good repeatability with correlation coefficient
  0.99

• Linearity 99.2

• Joint diameter computed from data is 62.8 mm,
  very close to measured diameter of 63 mm

13
Validation of OLE Sensing Modules

• On-body tests wearing OLE on human against
  BOPACs Goniometer and ShapeWrap from Measurand.
• Goniometer measures change of resistance in
  strain gauges
• ShapeWrap measure bend and twist angle of
  fiber-optic tape via light difference between
  inlet and outlet

14
Validation of OLE Sensing Modules

• Close relation of OLEs performance v.s.
  Goniometer and ShapeWrap.
• OLE able to handle high frequency excitation, but
  better with low frequency excitation
• Average RMS error
• OLE versus Goniometer is 3.8 with average
  correlation coefficient of 0.990
• OLE versus ShapeWrap is 3.1 with average
  correlation coefficient of 0.992.

normal flexion of 0.6 Hz
fast flexion of 2Hz
15
Benchmark with VICON

• Mean of angle difference µ -1.835
• Standard Deviation of Vicon and OLE s 3.332

16
Benchmark with VICON
17
Validation of SmartSuit system

• To examine repeatability and reliability of
  SmartSuit
• To testify SmartSuite as a complete arm motion
  capture system

Experimental procedure adapted from a therapy
section of stroke rehabilitation
18
Validation of SmartSuit system

• One data block file contains 10 trials of
  arm-reaching task
• 5 data blocks to produce 5 avg readings for each
  sensor
• Range and SD for each subject computed
• average range 2.819
• average standard deviation 0.697

19
Validation of SmartSuit system
ICC describes relative magnitude of two
components of the variability. Approximation of
ICC in short form is variability of
random errors is variability among their
average values computed over each repeated
measure As decreases, measurement
error explains a decreasing percentage of
variance in data, reliability increases, and ICC
approaches maximum value of 1. As
increases, measurement error explains an
increasing percentage of variance in data,
reliability decreases, and ICC approaches minimum
value of 0.
20
Validation of SmartSuit system

• ICC analysis performed for each sensor using
  Statistical Package for Social Sciences (SPSS)
• Average ICC for each sensor ranged from 0.959 to
  0.975
• Overall average 0.9670.08
• Average ICC is close to 1.00, indicating high
  reliability.
• High ICC values for all channels showing ability
  to perform and maintain its functions in routine
  circumstances, with different biometric subjects.

INTRA-CLASS CORRELATION COEFFICIENT OF
RELIABILITY
Shoulder Elbow Wrist (Bend) Wrist (Roll) Average
0.975 0.974 0.959 0.962 0.967
21
Conclusion

• Low cost high performance joint flexion sensor
• Patented optical encoding strip technology
• Accuracy of 0.1º and sampling rate of 1000Hz
• Flexible configuration and usages
• SmartSuit (Arm) based on OLE Accelerometer
• OLE sensor user validation
• SmartSuit user validation
• Clinical test bedding rehab application
  development

22
Thank You for Your Attention !
Team Members A/P Yeo Song Huat A/P Ling Keck
Voon Dr. Peter Luo Dr. Zhongqiang Ding Dr. Chee
Kian Lim Dr. Yan Liang
Funding support School of MAE, NTU ASTAR
SERC ASTAR EHS II Program ASTAR MedTech
Program ASTAR NKTH NRF IDM (MDA)
Collaborators A/P Henry Duh (NUS) Prof T-Y Li
(NCCU, TW) Prof M. Ceccarelli (Uni Cassino,
IT) Prof G. Stepan (BME, HG)
Wei-Ting Yang John Nguyen Kang Li Wei Ni Chao
Gu Ke Yen Tee
23
OLE Sensor Placement/Packaging
Shoulder packaging module
Sensor Placement
Shoulder packaging module
Wrist module
24
Human Arm Kinematic Model

• At shoulder, ?0?w0, and ?w1 are given by
  accelerometer of node 1

• Transformation from the world to frame 1

• So ?1 is calculated from

25
Human Arm Kinematic Model
At elbow, the joint angles ?2 and ?3 are given by
the OLE of node 1 and node 2 respectively.

• To find ?4, we expand transformation matrix from
  the world to frame 4

At wrist, the joint angles ?5 is given by the OLE
of node 3, ?w4 is ?w6 are from the accelerometer
of node 3.
Thus, expanding transformation matrix from the
world to frame 6 gives ?6