A REAL-TIME EMG PATTERN RECOGNITION CONTROL METHOD FOR ACTIVATION OF INSTRUMENTED WHEELCHAIR POWER ASSIST SYSTEM

Abstract

Pattern recognition control method is widely used for surface electromyography (sEMG) application to differentiate movement types according to Motor Unit Action Potential (MUAP). MUAPs detected from muscles are taken as indicators to activate DC motors in assistive equipment such as prosthetic hand and instrumented wheelchair. Performance of control method can be measured through classification accuracy and very important before commercialization. Therefore, the objective of this study is to measure the classification accuracy of pattern recognition control method classifier, which is called Probability Density Function (PDF), in predicting hand movement activities either in contact or recovery phase during wheelchair propulsion. Arduino board was designed to produce a command signal to activate the power-assist system (PAS) when the test subject is propelling the wheelchair forward. The developed method was tested against 5 able-bodied healthy subjects, where sEMG electrodes were placed on namely BIC, TRI, EXT, and FIX muscles. The accuracy results were found to be different for each subject. The highest was 99.4% while the lowest was 48.7%. It was found that low classification accuracy is due to PAS was activated in the recovery phase where it is supposed to remain in off condition. Consequently, PDF control method is effective for subject number 1 only where the hand movements have been successfully identified based on MUAP.