Learning-Based Classifying of sEMG Signals for Gait Event Detection Application

Recognition of the gait cycle plays an important role in rehabilitation especially in the field of lower limb rehabilitation robots that include prostheses and exoskeletons. In this paper, EMG signals, which are a tool for detecting user intentions, are used and FSR (Force-Sensing Resistor) data is assigned to them to classify gait cycle subphases. Then, after filtering the data and features extraction, learning-based methods including Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Artificial Neural Network(ANN), and Adaptive Neuro-Fuzzy Inference System (ANFIS) are utilized. MATLAB® was employed in order to construct mentioned classification methods. Results show that the ANFIS model yields better performance (۸۶.۸ percent accuracy) for the classification of the gait cycle swing and stance phases. Besides, SVM, KNN, and ANN accuracies are ۸۳.۵, ۷۸.۱, and ۷۹.۸ percent respectively. It is worth mentioning that in each application of gait analysis a trade-off between calculation cost and accuracy should be considered.

Keywords:

EMG signals , Gait Phase Detection , Wearable Device , Learning Classifiers , Gait Analysis