LAUSR

The purpose of current research work is to extract physiological information form surface electromyographic signal (sEMG) in efficient manner for different human locomotion and utilize it for lower limb prosthesis control. The proposed locomotion mode identification approach conserves the novelty in terms of its dependency on only single muscle electromyographic signal and independency on human gait phase. For current study, 18 healthy subjects of 21–42-year age group were engaged and their sEMG signal form two lower limb muscles has been recorded for three daily life locomotion’s. The presented approach of locomotion mode identification covers the wide group of designing factors. Here, twelve different window sizes, twelve types of feature vectors and six classifiers were compared on the ground of predictive performance and stability. The results show the best performance of overlapped windowing technique with window size of 256 ms and a shift of 32 ms. LDA emerges as best performing classifier (p value < 0.05) with a classification accuracy ranging from 89 to 99% for diverse feature subsets. Feature vector carrying time domain information reflected better performance. The multifactorial analysis reveals that the choice of feature vector as the most dominant source of performance variation (39.17% of total variance) and muscle selection as the least (1.35% of total variance). The proposed locomotion mode identification approach proves its applicability for rehabilitation and lower limb prosthesis control applications. Also, the protocol leads the researches for determining the appropriate values of designing factors involves in the model.