Intention Recognition Method for Sit-to-Stand and Stand-to-Sit from Electromyogram Signals for Overground Lower-Limb Rehabilitation Robots

Abstract

This paper presents a framework for classifying sit-to-stand and stand-to-sit from just two channel EMG signals taken from the left leg. Our proposed framework uses linear discriminant analysis (LDA) as the classifier and a multi-window feature extraction approach termed Consecutive Time-Windowed Feature Extraction (CTFE). We present the prelimnary results from 2 healthy subjects as a proof of concept. With the two tested subjects, we got predictive accuracies above 90%. The results show promise for a framework capable of recognizing the user's intention of sit-to-stand and stand-to-sit. Potential applications include rehabilitation robots for hemiparesis patients and exoskeleton control.