Auto-Segmentation Analysis of EMG Signal for Lifting Muscle Contraction Activities

Abstract

Time-frequency representation of a signal has been widely used in various research areas to analyze non-stationary signals (ie. electromyography (EMG) signals). However, due to the high computational complexity of certain time-frequency distribution techniques, the application of these techniques in the analysis of long duration EMG signals is not suitable. To overcome this problem, muscle contraction segmentation is essential to process the existed EMG signals, since not all of the EMG signal contains valid information to be analyzed. Thus, this paper presents an algorithm to automatically detect and segment the muscle contractions existed in EMG signal during long duration recordings. Surface EMG signals were collected from biceps branchii muscle of ten subjects during manual lifting. Subjects were required to lift a 5 kg load mass with lifting height of 75 cm until experiencing fatigue. The utilization of instantaneous energy of EMG is used to estimate the presence of first muscle contraction, second muscle contraction and until the last muscle contraction. This instantaneous energy is obtained from spectrogram and a threshold value is set to differentiate between muscle contractions and noise. This research shows that the algorithm is able to automatically segment muscle contractions in EMG signal based on the signal instantaneous energy.