The Effect of Signal Interval from sEMG sensors Combined with Mechanical Sensors on the Classification of Walking Patterns and Posture Activities

Forecasting human movement in various conditions is very challenging to control the exoskeleton robots for supporting daily life. This paper studied the proper time intervals posterior to the starting of movement to predict the type of posture. Surface electromyography (sEMG) signals were used as the marker of the movement starting time. Some characteristic features extracted from sEMG and mechanical signals were analyzed in classification. Four-time intervals posterior to the starting of the movement were studied in the experiments that included 500, 1000, 1500, and 2000 milliseconds intervals. The Ensemble Subspace Discriminant method was employed as the technique of classification. As the result, it was found that the combination between sEMG and mechanical signals achieved the best performance at 97.09±0.29% of accuracy. The classification results from all sensors processed at intervals 500, 1000, 1500, 2000 milliseconds with an accuracy of 86.21±0.49%, 92.9± 0.46%, 94.4±0.67%, and 97.09±0.29 %, respectively.