Wrist Movement Detection for Prosthesis Control using Surface EMG and Triaxial Accelerometer

The most important issue of prosthesis control is to get the correct control signal. In most studies, there is only one kind of signal applied to control the prosthesis, which is prone to error. In this study, a platform including measurement circuit and monitor software was developed to acquire mechanomyography (MMG) and electromyography (EMG) signals synchronously from flexor carpi radialis muscle of left arm as the signals to control prosthesis. The MMG signals were detected by a triaxial accelerometer, and they were analog preprocessed. The EMG signal was detected by three surface electrodes and an instrumentation amplifier was used to preprocess the differential EMG signal. For the first test, a pattern recognition experiment of four kinds of wrist movement was implemented. The experiment was carried out on six subjects. Using the Support Vector Machine (SVM) algorithm, the accuracy of pattern recognition classification was 96.06% by using MMG features combined with EMG features, which is higher than the accuracy of using just MMG (91.81%). The average accuracy of EMG features was 61.86%. It verified that acquisition of both the signals to control prosthesis would produce better results.