Abstract:
Mechanomyography (MMG) refers to the “sound” of muscle contracting, with frequency band from 2 to 100 Hz. MMG signal as a physiological signal source has been gradually utilized and justified in the control of prosthetic hands recently. This paper developed a way of constructing a forearm handmotion MMG feature space containing 18 time and frequency features,and principal component analysis (PCA) is adopted to reduce the feature dimensionality. Linear classifier algorithm is then applied to identify the four handmotion patterns (hand close, hand open, wrist flexion and wrist extension). Forearm handmotion MMG signals are acquired from 32 volunteers. The analysis results show that the average accuracy rate is above 95%, the recognition with threechannel acquisition configuration has the best overall performance, and the placement distribution of acquisition points on four forearm muscles has few effects on the accuracy rate.