Effective application of independent component analysis to the task of gradual EMG control

Abstract

In this work, the possibility of a reliable application of the independent component analysis to the problem of EMG interfaces is studied. The developed algorithm takes advantage of the experimental training procedure and enables a computationally efficient identification of gradual finger movements in real time by using inexpensive EMG sensors with a low sampling rate. Experimental data show that the algorithm improves the separation of motor units by 55 % as compared to raw EMG signals.