ON THE USE OF THE COMPLEX WAVELET TRANSFORM FOR SUBJECTS CLASSIFICATION AND CYCLE IDENTIFICATION FROM SEMG ANALYSIS DURING WALKING

摘要

This work proposes several innovative methods to group subjects and to identify the phases of the gait cycle applying the complex wavelet transform (CWT) to signals describing the electrical activity of muscles (EMG) during walking. The subjects classification is based on the extraction of the activation patterns (timing vs. level of activation) from selected wavelets coefficients. The cycle selection is achieved using preliminary physiological assumptions derived from experience about the muscles behavior during walking to process the wavelet coefficients. Two muscles (lateral Gastrocnemius and Tibialis anterior), three groups of subjects (asymptomatic adults, asymptomatic children and pathological children), and footswitch signals acquired concurrently with the EMG have been used to evaluate the results. The experiments demonstrated the prevalence of the results achieved for Gastrocnemius muscle. For both muscles, the asymptomatic adult group results to be discernible from the other two groups and the average absolute timing error between footswitches and EMG based gait cycle estimations results to be smaller than 100 ms.