Pattern recognition Of finger-motions based on diffusion maps and fuzzy K-nearest Neighbor classifier

Mechanomyographic (MMG) signal used for prosthetic hands has aroused the interest of a growing number of scholars in recent years and some considerable results have been achieved, however most MMG based approaches are limited to hand movements identification and control. In order to achieve a higher degree of freedom in hand movements, this paper proposed a novel method aiming at identifying the finger-motion patterns. Four-channel MMG signal was adopted to identify six single and combined finger-motion patterns. A total of 50 time-domain and frequency-domain features were extracted and diffusion maps were utilized to reduce the dimension of feature space. The fuzzy K-Nearest Neighbor (f-KNN) classifier was used to identify the six finger-motion patterns. The results showed that the average identification rate reaches to a high accuracy of 95.48 2.47%, which indicates that this algorithm is feasible and effective to identify the six finger-motion patterns and the MMG signal is a prospective alternative in the control of high freedom prosthetic hand.