p09027: upper extremity motion capture system -...

8
1 Abstract— This paper discusses the development and implementation of a wireless motion capture system for the human arm. The system is comprised of a mechanical brace with integrated rotary potentiometers. It also contains surface electrodes which acquire electromyographic signals through the CleveMed BioRadio device. The brace interfaces with a software subsystem which displays real time data signals. The software includes a 3D arm model which imitates the actual movement of a subject’s arm when under testing. Index Terms — BioRadio, graphical user interface (GUI), Qt, surface electromyography (sEMG) I. INTRODUCTION The goal of this project is to design a wireless biomechanical system which will be used for rehabilitation robotic research projects in the Biomechatronic Learning Laboratory (BLL) at the Rochester Institute of Technology. The system is being developed to study motion as it relates to the human robot interaction issues in rehabilitation. The specifications of the device were identified by the customer for the project, Dr. Edward Brown, the director of the BLL. Five main requirements were identified by him in the initial stages of the design of the project. These were: Four muscle groups required for measuring EMG activity Velocity measurement of limb movements Elbow angle, shoulder angle and angle between arm and body to be measurable Device should have 3 degrees of freedom Visual display of real time data with graphs The entire system is comprised of three functional blocks: mechanical, electrical, and software. The brace subassembly and platform subassembly form the mechanical block of the system. The brace is designed such that the electrical subsystem is integrated with it. The motion sensors identified by the team during the concept selection phase were rotary potentiometers which are integrated at the shoulder and elbow joints of the brace. Electric wires are run through the brace in an aesthetic manner which, connect to a printed circuit board (PCB). The CleveMed BioRadio device for measuring surface electromyographic (sEMG) signals connects to surface electrodes which are routed in a similar fashion as those associated with the rotary potentiometers and the PCB. A casing was selected and is placed on the platform subassembly at a specific location. The BioRadio is clipped on one side of the casing while the PCB and batteries are contained inside it. The software subsystem addresses the customer’s requirement for a visual display. Data is acquired from the BioRadio through a network interface and real time data is then displayed on a user friendly GUI. The complete system with a test subject strapped in is displayed below: Fig. 1: Complete System Depiction II. ELECTROMYOGRAPHY AND CONCEPT SELECTION A. Electromyography (EMG) The evaluation and recording of muscle activity is accomplished through a technique known as electromyography (EMG). In order to quantify EMG, it is imperative to understand the concept of a motor unit and the associated impulse or action potential generated when the motor unit fires. A motor unit is composed of a motor neuron and the muscle fibers it further stimulates. Once the motor unit fires, the action potential is transmitted across the neuromuscular junction upon which an action potential is generated in each of the muscle fibers connected with the motor unit. A motor unit action potential (MUAP) is the sum of this electrical activity. Fig. 2 below depicts a motor unit. EMG enables the study of electrophysiological activity from a number of motor units collectively. P09027: Upper Extremity Motion Capture System Melissa Gilbert, J.J. Guerrette, Dan Chapman, Alan Smith, Adey Gebregiorgis, Pooja Nanda

Upload: hatuyen

Post on 01-Jul-2018

214 views

Category:

Documents


0 download

TRANSCRIPT

1

Abstract— This paper discusses the development and

implementation of a wireless motion capture system for

the human arm. The system is comprised of a mechanical

brace with integrated rotary potentiometers. It also

contains surface electrodes which acquire

electromyographic signals through the CleveMed

BioRadio device. The brace interfaces with a software

subsystem which displays real time data signals. The

software includes a 3D arm model which imitates the

actual movement of a subject’s arm when under testing.

Index Terms — BioRadio, graphical user interface (GUI), Qt,

surface electromyography (sEMG)

I. INTRODUCTION

The goal of this project is to design a wireless

biomechanical system which will be used for rehabilitation

robotic research projects in the Biomechatronic Learning

Laboratory (BLL) at the Rochester Institute of Technology.

The system is being developed to study motion as it relates to

the human robot interaction issues in rehabilitation.

The specifications of the device were identified by the

customer for the project, Dr. Edward Brown, the director of

the BLL. Five main requirements were identified by him in the

initial stages of the design of the project. These were:

• Four muscle groups required for measuring EMG

activity

• Velocity measurement of limb movements

• Elbow angle, shoulder angle and angle between

arm and body to be measurable

• Device should have 3 degrees of freedom

• Visual display of real time data with graphs

The entire system is comprised of three functional blocks:

mechanical, electrical, and software. The brace subassembly

and platform subassembly form the mechanical block of the

system. The brace is designed such that the electrical

subsystem is integrated with it. The motion sensors identified

by the team during the concept selection phase were rotary

potentiometers which are integrated at the shoulder and elbow

joints of the brace. Electric wires are run through the brace in

an aesthetic manner which, connect to a printed circuit board

(PCB). The CleveMed BioRadio device for measuring surface

electromyographic (sEMG) signals connects to surface

electrodes which are routed in a similar fashion as those

associated with the rotary potentiometers and the PCB. A

casing was selected and is placed on the platform subassembly

at a specific location. The BioRadio is clipped on one side of

the casing while the PCB and batteries are contained inside it.

The software subsystem addresses the customer’s

requirement for a visual display. Data is acquired from the

BioRadio through a network interface and real time data is

then displayed on a user friendly GUI.

The complete system with a test subject strapped in is

displayed below:

Fig. 1: Complete System Depiction

II. ELECTROMYOGRAPHY AND CONCEPT SELECTION

A. Electromyography (EMG)

The evaluation and recording of muscle activity is

accomplished through a technique known as

electromyography (EMG). In order to quantify EMG, it is

imperative to understand the concept of a motor unit and the

associated impulse or action potential generated when the

motor unit fires.

A motor unit is composed of a motor neuron and the muscle

fibers it further stimulates. Once the motor unit fires, the

action potential is transmitted across the neuromuscular

junction upon which an action potential is generated in each of

the muscle fibers connected with the motor unit. A motor unit

action potential (MUAP) is the sum of this electrical activity.

Fig. 2 below depicts a motor unit. EMG enables the study of

electrophysiological activity from a number of motor units

collectively.

P09027: Upper Extremity Motion Capture

System

Melissa Gilbert, J.J. Guerrette, Dan Chapman, Alan Smith, Adey Gebregiorgis, Pooja Nanda

2

Figure 2: Visual Illustration of Motor Unit and Motor

Neuron

Source: www.abcbodybuilding.com

Two different types of EMG are commonly used:

intramuscular and surface. Intramuscular EMG is invasive and

for this particular system, surface EMG was more suitable.

Surface electrodes provide a general picture of muscle

activation which is more useful for physiotherapy.

A number of different muscle groups [2] were researched to

understand which would be the most important for us to target

in accordance with the goals and objectives of the project.

Upper arm movement is governed by the deltoid while the

primary muscles for extension and flexion of the lower arm

are the triceps and biceps. Lastly, the brachioradialis was

chosen to be monitored as it assists the biceps and triceps

muscles for movement of the forearm. A visual illustration of

these four muscle groups is provided in Fig. 3 below.

The left upper corner figure depicts the biceps brachii while

the right upper corner figure depicts the triceps brachii. The

left lower corner shows the deltoid muscles and the right

lower corner is the brachioradialis.

Since another customer requirement was the capability of

velocity and time measurements of different limb movements,

we had to consider a number of motion sensors and choose the

most appropriate one.

Figure 3: Visual Illustration of the biceps brachii,

triceps brachii, brachioradialis and deltoid muscles in

clockwise order

B. Motion Sensors

We looked at several different motion sensors such as

accelerometers, gyroscopes, optical encoders, potentiometers.

A concept selection matrix was developed to scope the

advantages and disadvantages of each. The selection criteria

we used were:

• Signal Processing

• Mechanical Integration

• Power Requirement

• Position Accuracy

• System Overhead

• Cost

Each criterion had a different weight of importance

associated with it. Considering our specific goals, signal

processing, mechanical integration and position accuracy were

our most important criteria. We rated each motion sensor

based on information we researched and the sensor with the

highest score was chosen. As per our selection matrix,

potentiometers ranked the highest.

Once we narrowed down our motion sensor to

potentiometers, the next step was to decide which type of a

potentiometer would be used. For this, we ordered samples of

different types of potentiometers and performed some

engineering analysis. In essence we wanted to understand the

error in measurement from each potentiometer once integrated

with the brace. From this analysis, we came to the conclusion

that rotary potentiometers would be the best way to go for our

project. The rotary potentiometers we are using are single-turn

conductive plastic with a resistance of 10k and +/-1% linearity

tolerance with a life expectancy of 10 million cycles [3]. The

price of these potentiometers is economical and they are easily

available such that a replacement can be done if necessary.

III. DESIGN AND DEVELOPMENT

A. Electrical Circuit

The motion sensors require a constant voltage supply at all

times which is accomplished using two voltage regulators. A

9V battery powers the circuit and the voltage regulators reduce

the 9V signal to 2V and 4V signals [4]. The CleveMed

BioRadio takes in differential inputs limited to a range of 2V

in the positive or negative direction [5]. In order to fully

utilize the input ranges of the BioRadio, the potentiometers are

powered by 4V allowing for the wipers of each potentiometer

to swing from 0 to 4V. One differential input of the BioRadio

is then connected to the wiper of the potentiometer, while the

other input is connected to a virtual ground which is created

by the 2V offset coming from the second voltage regulator.

This setup ensures that as long as the potentiometer’s wiper

varies between 0 to 4V as desired, the input of the BioRadio

will read anywhere between -2V and +2V.

The designed circuit schematic is shown in Fig. 4. From this

circuit schematic, a PCB layout was created and the

dimensions of the board were determined to be 1” x 1”. This

layout is shown in Fig. 5.

3

B. Mechanical Brace

The brace subassembly is designed to fit the 95th

percentile

of both the female and male population [1]. The basic

components of the brace are the lower arm and upper arm

sections with the hinged elbow joint to lock them together, the

hand grip attached to the lower arm section, foam padding on

the lower arm which was added for safety and comfort of the

test subject, and lastly, the 2-axis shoulder joint. The three

rotary potentiometers are integrated such that one fits at the

hinged elbow joint, while the other two fit at the shoulder

joint. A Solid Works model of the brace exhibiting the

integration of the rotary potentiometers is shown in Fig. 6

below.

The assembly of the entire brace is complex and can be

broken down into smaller subassemblies. The first step is to

put together the handle subassembly or the hand grip. There is

a main hand back plate with side plates on both ends

supporting the hand of the subject. A handle is incorporated

through the side plates and essentially, a subject will be

holding on to this handle when using the brace. The handle

has the provision for attaching ringed weights to it as the

customer desired to have the possibility of performing

experiments with the user carrying weights as well.

The next step for the brace development is the elbow joint

assembly. Two plates have been designed for the lower arm,

one which has the channel insert and the other which can be

inserted. Similarly there are two plates for the upper arm. This

type of design allows for length adjustability which is a main

criterion for the brace as it should fit 95th

percentile of the

population as mentioned earlier. The upper arm plate with the

channel insert and the lower arm plate without the insert attach

to form the elbow joint. At the elbow joint area, a

potentiometer bracket is attached and this is where the

potentiometer gets integrated with the brace.

The next assembly is the lower shoulder joint assembly.

The free end of the upper arm plate without the channel insert

is connected to the lower shoulder side plate and a

potentiometer bracket is attached here as well. This is the

second location for potentiometer integration. Lastly, the

upper shoulder joint assembly is considered. Two side plates

connect to a brace beam, which is a square tube designed as

the connection to the platform. On the side of the brace beam,

the third potentiometer bracket is attached and the open ends

of the side plates are inserted into the lower shoulder joint

assembly.

C. Mechanical Platform

An additional requirement of our customer was that a

subject could undergo testing while sitting or standing. In

order to have these conditions fulfilled, a platform was

suggested to which the brace could be attached. During the

concept development phase, some basic requirements of the

platform were identified to be height adjustability, portability,

compatibility with a wheelchair as well as providing left and

right hand compatibility. The actual designed platform has all

the above mentioned requirements along with three linear

adjustments and appropriate shoulder pivot height range for

both the standing and sitting situations. The Solid Works

model created for the platform is illustrated below in Fig. 7.

Figure 4: Circuit Schematic depicting all three

potentiometers, the two voltage regulators and the

BioRadio connections

Figure 5: PCB Layout

Figure 6: Brace subassembly highlighting the major

components

The H-shaped metallic base of the platform is 66”x38.5”

which is big enough to accommodate a wheelchair as desired.

The diagonal flat tube connections in the lower half of the

platform ensure the stability and sturdiness of the platform.

Standard sized bolts, nuts and washers are used to connect the

different components of the platform together. For height

adjustability, the poles attached to the base have hollow

channels in them. On each side a second pole tube is inserted

in the base poles and with a handle, the desired height can be

set.

4

The two horizontal bars which connect the pole inserts

accomplish the dual functions of sitting and standing. If the

test subject is sitting, the brace is connected through the brace

beam to a brace integration T which is set on the lower

horizontal bar. Another brace integration T is set on the upper

horizontal bar. For an experiment where the test subject is to

be standing, the brace along with the beam just has to be

shifted and attached to the T on the upper bar.

The platform also has wheels on the bottom for smooth

transportation and movement if necessary. The total

assembling time of the platform is roughly an hour.

Figure 7: Platform Subassembly Model

D. Software Subsystem

The main elements of the software subsystem are the

hardware interface between the BioRadio and the computer

software, the graph modules which graph the data output from

the BioRadio, the mathematical arm model which will be a

visual rendering of the actual human arm under test and lastly

the Graphical User Interface (GUI) which displays all the

different graphs as well as the different menu options for the

user to process or display data the way he or she desires. All

the software coding has been done using C++ and the

development has been done in a cross-platform application

framework called Qt. Fig. 8 below shows the class diagram

for the entire software subsystem.

Figure 8: Software Class Diagram

The development of the software was done in four different

stages. The first stage was to create a software – hardware

interface which essentially is provided by CleveMed through

the libraries in the software development kit. The next stage

was to learn how to acquire data from the BioRadio and

display the data in graphical form. This required the creation

of graph modules. Originally, graph modules were created in

another platform called Open GL. However, when integration

with Qt was attempted, the integration was not successful. In

order to overcome this obstacle, the graph modules were

recreated in Qt.

One additional requirement from the user was to see the

corresponding angle of the test subject’s movement on the

GUI. Hence, the third stage was to develop an arm model

which could render to a 3-D matrix array to depict the angles

of the brace movement. The 3-D design of the human arm is

obtained using a mathematical arm model consisting of

rotation transformation matrices. The model uses the different

angles measured by the potentiometers as its input variables

and outputs the elbow and wrist coordinates in Cartesian form.

Fig. 9 below shows the brace arm model with the

corresponding joint angles:

Figure 9: Illustration of brace arm model with Angles

converted to 3-D Arm Model in the GUI

The original coordinates of the elbow are:

[ ]0 0 upper arm lengthT

− (1)

while the original coordinates of the wrist are:

[ ]0 0 upper arm length+lower arm lengthT

− (2)

The transformation matrices used for the elbow and wrist

respectively are:

(3)

(4)

5

Using these matrices and coordinates, an algorithm for

calculating the new coordinates is incorporated in the software

code.

The final stage for the software development was to

program the actual GUI. Using Qt, all the menus, the different

buttons and the data displays are developed. The components

of the menu bar are as follows:

• New Session – allows the user to create a new

session

• Open Session – allowing the user to open an existing

session

• Channel Window – user gets to select the channels to

be overlaid on the desired channel window

• Animation – allows the user to see the animation of

the arm model

• Filters – allows the user to select a filter to process

the raw data

Also, the GUI has buttons which are described as follows:

• Change Capture – replay data saved before

• Start – enables user to start collecting data from the

channels being graphed

• Stop – ends recording of data

• Mark – allows the user to set a marker at any location

on the data

The graphs are integrated into the GUI such that a user can

view data collected from all 8 channels of the BioRadio. These

graphs can be overlaid if someone wants to see the motion and

muscle data together.

IV. TESTING AND EVALUATION

To ensure that we had met all the requirements and

specifications of the project, each system was tested

independently first to evaluate its functionality. Once the

performance of the individual systems was verified,

comprehensive system testing was done to ensure there were

no hiccups upon integration.

A. Mechanical Testing

After the brace was assembled, the first thing tested was the

smoothness of the joints. A key aspect for accurate motion

capture was that the joints of the brace did not exhibit

jerkiness. When the brace was designed, bearing spacers were

included for the purpose of ensuring smooth joint operation.

The brace was fitted to everyone on the team and then tested

with several flexing and extending movements to verify the

smooth operation of the joints. The motion data being

captured and displayed on the GUI was monitored and no

blips in the signals were observed.

Another important specification for the brace was comfort

when a subject was wearing it for a considerable amount of

time. Three members of the team underwent comprehensive

testing for an hour at a stretch with the brace strapped to their

arm and there was no issue of pinching or tightness of straps.

The only point that came up during this testing was the

thickness of the foam pads. Ideally, the foam pads should be

thinner to provide more comfort to a test subject and this will

be addressed in a future redesign.

Since the brace was designed to fit a large sample

population, we needed to test the length adjustability of the

brace at least on the 6 of us within the team and another 4

individuals outside the team to ensure the brace was

compatible with different arm lengths and sizes. As per our

testing the brace fit a number of people with different arm

lengths and sizes; signifying the brace was designed within the

specified range.

Another important test was to see if the platform was stable

enough. When a subject was undergoing tests, we needed to

evaluate if the subject’s movement was affecting the stability

of the platform. After the platform was assembled, it was

vigorously shaken to observe if it bounced around a lot. Our

preliminary results were that it did not bounce too much and

was stable enough to ensure accurate operation of the system.

We also tested stability by observing the movement of the

platform, if any, when a test subject was strapped in and

moved his/her arms around vigorously and in all directions.

Once again, there was no corruption of data or movement of

the platform along with the test subject.

The final testing done on the platform was a weight test. The

customer stated at the beginning of the project that the system

should be portable and mobile. This implied that the platform

designed had to be easily disassembled and then transported

from one place to another. The design of the platform is such

that only 4 bolts need to be removed to divide it into two

easily transportable pieces; the base with wheels and the upper

assembly which can be carried. The team tested this out by

transporting the platform from the lower level of the building

to the third level and within ten minutes the platform was

taken apart and the two pieces were easily carried to an

elevator and transported.

B. Software Testing

The main testing for the software was in regards to the GUI

and the code linking the hardware to the software. The first

test conducted was verifying that the BioConnect code (a code

provided by CleveMed systems) executed without errors and

that data was being received from the bioinstrumentation

device. The BioRadio was set-up with the channel tester and

all eight channels were programmed to be on. The code was

then executed and as expected, data was received from the

BioRadio.

Another important test was the display speed of the data. A

test subject was strapped in the brace and main program was

executed. The test subject was asked to move his arm in

different directions and no delay between the subject’s

movement and data display in the GUI was noticeable.

In order to test the GUI, the first step was to ensure that

when the main program was executed, no error was obtained

and the program would not quit until the user decided to close

it. Initially there were some issues with the execution of the

main program. After a number of changes and refinement of

code, the main program execution errors were addressed and

the program now runs smoothly without any issues.

The GUI was designed to provide the user with several

options for processing data, saving it, calibrating the system,

using the help menu and so on. Some options are a specific

component of the menu bar while others are front window

buttons on the GUI. Each component on the menu bar has

been tested thoroughly by clicking several times and making

sure that the user is directed to the correct link.

6

Simultaneously, each button on the GUI has been clicked as

well and proper actions are completed upon every selection.

It was also needed to ensure that the graphs are being

displayed with the correct data from the appropriate channel.

In depth testing has been done to verify that the window data

displays accurately and is properly resized. The GUI has the

ability to overlay data from different channels as per the user’s

requirements.

One last test for the software that was undertaken was to

ensure that the mathematical arm model replicated the actual

movement of the brace and was at the same angle as the brace.

C. Electrical Testing

A circuit test was done to verify that the printed circuit

board (PCB) was operating as expected. Voltages were

measured at a number of different vias and the measured data

was within the acceptable range of the expected values. The

PCB had been designed and built correctly.

Another important test was the dead-zone test. Essentially

each rotary potentiometer has a dead-zone and it was

necessary for accurate system operation that the dead-zones

would be in a region where no measurements would be taken.

This required extremely accurate machining during

mechanical manufacture of the brace. To test the location of

the dead-zones, the brace was set-up and each joint was

rotated to the extremes and the voltage at that point was noted

using the GUI. It was verified that the voltage at each

extremity was either below 2V or above -2V and hence the

location of the dead-zones was accurate.

D. System Testing

A life span test was also conducted to measure the battery

life of the system. The original specification was that the

system should be kept on for more than 2 hours and there

should be no change of batteries or crashing of the system. In

order to verify this, the system was kept on for more than 4

hours. No battery change was required and the software did

not crash either.

The most important system test was the angle/calibration

test. Each joint had to be calibrated using the calibrate

function in the GUI and a goniometer had to be used to

manually measure the angles to be fed into the software.

Three test subjects were used for system calibration and the

process was done using 2 points first and then 3 points. The

angles measured are compared with those given by the

software. Once calibration is done, the test subject is required

to move the joint under consideration at different angles and

the angles given by the software are recorded along with the

manually measured angles. The results obtained for the

calibration testing are summarized in the Results section.

V. RESULTS

Overall, the entire system is functional and each individual

system performs as desired. Table I gives a summary of the

desired and actual values of different specifications for the

entire mechanical system.

Table II gives a summary of the software system

specifications while Fig. 10 shows a screen shot capture of the

functioning GUI demonstrating overlaying of angle data on

EMG data in all three channel windows. We verified that the

mathematical arm model moves in accordance with the brace

and is accurate as far as angular position is concerned which

was a requirement of the customer. Fig. 11 depicts a screen

shot of the mathematical arm model.

Description Desired Actual

Shoulder height

range (standing)

528 – 674

mm 47.5 - 58.5 in

Hand width range 81 – 110 mm 5 in

Shoulder rotation range

Transverse plane -60° to 140° << -60° to 53°

Vertical plane -70° to 190° - 38° to 140°

Elbow rotation range 0° to 160° 0° to 163°

Weight of brace 5 lb 4.4 lb

Comfortable padding Yes yes

Table I: Summary of Mechanical System Specifications

Description Desired Actual

Time delay from actual

movement to movement

in software 100 ms Less than 100 ms

OS compatibility WinXP All

Disk space 2 GB 500 MB

Processor speed 2 GHz 3GHz

Signals displayed 9 21

Animation Yes No

Table II: Summary of Software System Specifications

Fig. 10: Screen Shot of GUI Demonstrating Angle Data

Overlaid on EMG data

7

Fig. 11: Screen shot of mathematical arm model

animation

As mentioned earlier, the testing done regarding the PCB

was successful and Table III below gives a comparison of the

expected and actual values. Table IV summarizes the

specification of the electrical system.

Expected

(V) Actual (V) % Error

Battery 9.000 8.814 2.07%

U1 Output 2.000 1.978 1.10%

U2 Output 4.000 4.013 0.31%

Virtual

GND 2.000 1.996 0.20%

Table III: Expected and Actual Voltage Values for PCB

Testing

In order to summarize the testing conducted for system

calibration, the average error for each joint for each test

subject was plotted for both the 2-point and 3-point calibration

processes. Figures 12 – 14 below depict the results of the

calibration testing.

Fig. 12: Plot of Elbow Joint Calibration Testing

Fig. 13: Plot of Shoulder 1 Joint Calibration Testing

Fig. 14: Plot of Shoulder 2 Joint Calibration Testing

Description Desired Actual

Accuracy of angular position

Elbow Less than ±1° ±5°

Shoulder Less than ±1° ±5°

Accuracy of angular velocity

Elbow Less than

±2.5°/sec DNE

Shoulder Less than

±2.5°/sec DNE

Data acquisition

frequency 960 Hz 960 Hz

No. muscle groups

monitored At least 4 4

Life cycle of motion

sensors Greater than 1

million 10 million

Table IV: Summary of Electrical System Specifications

8

VI. DISCUSSIONS AND CONCLUSIONS

On the whole, the product delivered to the customer was

almost on par with his requirements and specifications. The

team had issues meeting his specification regarding the arm

model integration with software due to time constraints.

Besides that hiccup, the system is robust and functional and

the entire project was well under the budget which was a

major success milestone.

A future generation of the project could most definitely be

conceived and one major improvement to the system would be

to design a bioinstrumentation device instead of using the

device from CleveMed systems. This would make the system

completely independent.

VII. REFERENCES

[1] A. Tilley and H. Dreyfuss, “The Measure of Man and

Woman: Human Factors in Design”, Wiley Publications,

2002

[2] A. Adams. “The Muscular System.” Human Body

Systems. Michael Windeslpecjt. Series Editor. Greewood

Press, Westport, Connecticut, London. 2004.

[3] P3 America, Inc: Single-Turn Conductive Plastic

Potentiometer Series – MP20, 21

[4] National Semiconductors. LP2989LV Micropower 500

mA Low Noise Low Dropout Regulator for Output Voltages.

June 2004. http://cache.national.com/ds/LP/LP2985LV.pdf.

[5] BioRadio 150 User’s Guide, Cleveland Medical Devices

[6] K. Itoh, H. Miwa, Y. Nukariya et al., “Development of a

Bioinstrumentation System in the Interaction between a

Human and a Robot”. Proceedings of the 2006 IEEE/RSJ Intl.

Conf. on Intelligent Robots and Systems, Oct 9 -15, 2006,

Beijing, China