Evaluating the Performance of Neural Network and Kalman Filter Based Linear Model on Classification of Hand EMG Signals

In recent years, many revolutionary algorithms were designed for enhancing the performance of the neural network classification. This paper aims at evaluating the efficiency of one of these algorithms in intuitive control of the prosthetic hands. We used a combination of a neural network and a Kalman filter based linear model for the classification of 4 movement patterns by recruiting a single electromyographic channel electrode. The resultant recognition accuracy reached 95.4% with a mean squared error of 0.0473. The results show that the proposed technique is promising and competitive compared to traditional classification strategies.