USING GYROSCOPE AND ACCELEROMETER IN THE FES HAND GRASP SYSTEM

K.Y. Tong, A.F.T. MAK

Jockey Club Rehabilitation Engineering Centre

Hong Kong Polytechnic University

Hung Hom, Kowloon, Hong Kong

ABSTRACT

Restoration of a few basic upper extremity functions to individuals with cervical spinal cord injury would significantly increase the independence and improve their quality of living. Different FES systems have been used to restore the hand functions. These individuals had limited voluntary movement, a reliable control system should be able to distinguish when the person would like to turn on the stimulation and select different stimulation patterns from their normal daily activities.

A new control strategy has been designed to identify a pre-defined movement (move forward and then backward) using accelerometer and gyroscope attached on the shoulder and upper arm position in order to trigger the ON/OFF event. Five normal subjects were tested with the control strategy using a real-time software and the results showed this control strategy was very reliable (100% successful rate to trigger the ON/OFF event within a 3-second period) and easy to control. This control technique provides the subject to have the freedom to move and achieve different tasks without accidentally activated the ON/OFF event. Only the pre-defined movement could activate the stimulation. The tilt angle of the body segment was measured by the accelerometer, which could provide extra information for the FES system to deliver different stimulation patterns.

INTRODUCTION

Functional electrical stimulation (FES) was first used in 1961 by Liberson to stimulate the paralysed muscle to resolve the drop-foot problem of hemiplegics after stroke. In the past three decades, different researches have been conducted to restore the functional movement of the paralysed extremities [1]. The greatest improvement in the quality of life of individuals who have suffered cervical spinal cord injury lies in the restoration of some primitive hand functions like pinch and grasp. Different FES systems have been developed to restore hand functions in spinal cord injured (SCI) persons through electrical stimulation [1,2]. Safety and reliability are the important issues of the FES control system. These paralyzed individuals had limited voluntary movement (e.g. shoulder movement), a reliable control system should be able to distinguish when the person would like to turn on the stimulation and select different stimulation patterns from their normal daily activities. The system must provide good performance and easy to don/doff. The most common systems used a simple pre-programmed stimulator with an on/off switch to control the stimulation. Some advanced systems used a joystick-like sensor to record the shoulder elevation and wrist extension to control opening and closing of the hand [2]. Besides, the "natural" signals recorded from the user's body have also been used to control the stimulation [3].

Nowadays, FES becomes more mature with the latest technology and the systems can pass the barrier of the restriction of technology. Recently accelerometer and gyroscope have been investigated to detect gait events for the FES walking system [4]. These sensors are compact, lightweight, easy to don/doff and require only a single contact point on the skin. We have used these sensors to record the arm and shoulder movements, and a new control strategy using a pre-defined movement has been designed to capture the subject intention for controlling the stimulation on the upper extremity. This study evaluated the performance of the control strategy using accelerometer and gyroscope around the shoulder and elbow position.

METHODS

Movements are classified into two catalogues: linear and angular. In a normal body movement, if a body segment is moved from position A to position B (A®B), it first accelerates from the start position and when it is near to the target position it starts to decelerate. The displacement signal likes the sigmoid curve in figure 1a. The velocity graph shows a bell-shape in figure 1b and the acceleration graph shows a peak and a valley in figure 1c. If the body segment moves from position A to position B and then immediately back to A (A®B®A, move forward and then backward) (figure 1d). The signal will provide one peak and one valley in the velocity graph (figure 1e), and two peaks and one valley in the acceleration graph (figure 1f). With these features, a threshold can be set to detect the peak and valley and used the time delay between the valley and peak to distinguish the A®B®A movement pattern from the normal movement (A®B or B®A).

c

b

a

A®B

Displacement Velocity Acceleration

A®B®A

f

e

d

Figure 1 : Displacement, velocity and acceleration graph for the A®B and A®B®A movement.

Accelerometer (ADXL202, Analog Device, USA) and gyroscopes (ENC-05D, MuRata, Japan) were used in the experiments. The on/off event for the FES system was triggered when the sensor detected the A®B®A movement pattern. The accelerometer was used to measure the linear acceleration. The two-peak feature in the figure 1f was used to identify the on/off event. If more than one peak were above a prescribed threshold within a time period, it would trigger the on/off event. The gyroscope was used to measure the angular velocity. The on/off event was identified using the one peak and one valley feature in figure 1e. If the control system detected a peak and a valley above a threshold within a time period, it would trigger the on/off event.

Five normal subjects (average age 27) were recruited to evaluate the control strategy and no FES was applied during the experiment. Software was written in Labview (National Instruments, USA) to detect the on/off event from the sensory signals in a real-time mode. A visual and audio feedback was given to acknowledge the subject when an on/off event was detected. Three sets of test were conducted as following:

Figure 2: Upper arm movement (a) resting position, and (b) raise up the elbow

Accelerometer on the upper arm

Subject sat in front of the computer with the elbow flexed at 90° in the resting position (position A, figure 2a). The elbow could easily move up to position B (figure 2b), and this movement included an internal rotation and abduction of the arm. The hand was rested on the table all the time when the elbow was moved up and down. Then we used the “move forward and then backward” (A® B® A) pattern for triggering the On/Off event. Sensors were attached on the skin surface using self-adhesive tape or elastic strap. The accelerometer was attached on the lateral aspect of the upper arm (5cm above the lateral epicondyle of humerus, figure 2), the sensor axis was pointing towards the lateral direction of the arm.

The experiment was conducted in three parts. In the first part the subject was asked to trigger the On/Off event and rested for 5s after each event. This procedure was repeated 20 times. If the subject failed to trigger the event in the first attempt, they tried the second and third attempt until the event was successfully triggered or stopped after a 3-second period from the first attempt. The second part was to use the accelerometer to measure the tilt angle of the upper arm, The subject was asked to use the tilt angle (accelerometer signal) to follow a target signals (square wave, 6s cycle time). The subject traced 20 square waves and the mean squared error was calculated. The third part was to merge the first two parts together. In one cycle, the subject first needed to trigger the On/Off event and then traced the target signals (two square waves, each 6s cycle time) and then triggered the On/Off event and rested for 12 seconds before starting another cycle. This cycle was repeated five times.

Accelerometer on the shoulder

The accelerometer was attached on the clavicle to detect the shrug movement on the shoulder, and the sensory axis was pointing the vertical-up direction. The first part of the previous test was conducted (trigger the on/off event) on all the five subjects. However, the tilt angle of the shoulder was too small, and the subjects were unable to trace the target signal. Therefore the second and the third part of the previous test could not be conducted.

Gyroscope on the upper arm

Gyroscope was attached above the elbow position (figure 2) to measure the angular velocity of the internal rotation of the arm, and the sensory axis was paralleled to the longitude axis of the humerus. A subject was recruited to evaluate the performance of using the angular velocity to trigger the On/Off event and the “trigger the on/off event” test was conducted.

RESULTS

b

a

The waveforms of the accelerometer and gyroscope during the “move forward and then backward” (A®B®A) movement are showed in figure 3.

Figure 3 : The signals of the A®B®A movement were measured by the gyroscope and the accelerometer on the upper arm are shown in (a) and (b) respectively

Five normal subjects had been tested with the accelerometer and the results showed this control strategy was very reliable (100% successful rate to trigger the ON/OFF event within a 3-second period). The results showed the system are easy to control and on average less than two attempts were required The results showed the sensor attached on the upper arm had better control performance than on the shoulder. NRMSE is the normalised root mean square error (RMSE / amplitude of the target signal).

In third part of the upper arm test (trigger + Trace), the results showed that the signal during tracing the target line and the signal for triggering the on/off event could be distinguished, and no on/off event was activated during the trace period. Moreover, the average number of attempt to trigger the On/Off event of the gyroscope on the upper arm was 1.05.

DISCUSSION

This control technique provides the subject to have the freedom to move and achieve different tasks without activated the stimulation ON/OFF event. Only the pre-defined movement could activate the stimulation. The tilt angle also can been measured by the accelerometer on the upper arm position, which can be used to instruct the system to provide different stimulation patterns. The results showed a good potential to use gyroscope and accelerometer to capture the subject intentions. A further investigation will be conducted on SCI persons to evaluate this control strategy.

REFERENCES

[1] Teeter JO, Kantor C and Brown D (1995), Functional Electrical Stimulation(FES) resource guide for persons with spinal cord injury or multiple sclerosis, Cleveland FES Center

[2] Mulcahey MJ, Betz RR et al., (1997), "Implanted functional electrical stimulation hand system in adolescents with spinal injuries: an evaluation", Arch Phys Med Rehab, Vol 78, pp597-607

[3] Graupe D (1989), EMG pattern analysis for patient-responsive control of FES in paraplegics for walker-supported walking, IEEE Tans. Biomed. Eng., Vol. 36, No. 7, pp 711-719

[4] Willemsen AT, Bloemhof F, Boom HB.(1990) Automatic stance-swing phase detection from accelerometer data for peroneal nerve stimulation. IEEE Trans. Biomed. Eng., Vol.37 pp1201-1208