A Hybrid WD-EEMD sEMG Feature Extraction Technique for Lower Limb Activity Recognition

Classification and analysis of surface EMG (sEMG) signals have been particular interest due to their numerous applications in the biomedical field. They can be used for diagnosis neuromuscular diseases, kinesiological studies, human-machine interaction. However, these are difficult process noisy nature. To overcome this problem, a hybrid wavelet with ensemble empirical mode decomposition pre-processing technique called WD-EEMD is proposed classifying lower limb activities based on sEMG healthy knee abnormal subjects. First, Wavelet De-noising filtering out white Gaussian Noise (WGN) unwanted (contribution other muscle signals). Next, an Ensemble Empirical Mode Decomposition power line interference (PLI) baseline wandering (BW) noises, followed by extraction total nine time-domain features. Finally, performance parameters Linear Discriminant Analysis (LDA) classifier calculated 3-fold cross-validation technique. This study involves 11 individuals abnormality three different activities: walking, flexion leg up (standing), extension from sitting position (sitting). Different techniques similar that were compared. It was observed method achieves average classification accuracy 90.69% 97.45% subjects subjects, respectively.