Classification of ADLs using muscle activation waveform versus thirteen EMG features

Abstract

Movement classification has been a challenging problem in neuroprosthesis control. Many studies have taken into account the classification of movement using time and frequency domain features extracted from the electromyogram signals while calculating these features are usually time consuming. In this paper, we compared the capability of muscle activation waveform in the classification of five arm movements during activities of daily living, also known as ADLs, versus 13 different prevalent electromyogram features. We tested our technique on the electromyogram signal recorded from six healthy male right handed subjects. We, also, selected the muscles that are supposed to be the intact muscles in a tetraplegic spinal cord injury patient. Our results indicated that there exists significant higher accuracy with recruiting muscle activation waveform in classification, while the complexity of calculating features is eliminated.