Feasibility of an Intelligent Home-Based Neurorehabilitation System for Upper Extremity Mobility Assessment

Abstract

Health personnel are often unavailable for supervised robot-Aided neurorehabilitation in hospitals, and patients are usually challenged by transportation issues to get to the hospital. Thus, a discontinuity between therapy in the hospital and at home appears slowing down the upper extremity (UE) mobility recovery. The aim of this work was to develop a system, based on wearable devices and EMG armbands, able to assess the quality of the UE joint movements and intelligently guide the patients during home-based rehabilitation. This system fuses a classification model together with a dynamic time-warping (DTW) analysis. The classification model was trained with UE joint movements gathered from clinicians, obtaining more than 80% accuracy using only five joint angles. Then, the system was tested on two poststroke patients and a healthy subject. The results suggest that the proposed system can be: 1) a useful tool for clinicians to evaluate the rehabilitation therapy and 2) an intelligent system able to make decisions based on the quality of the activity executed at home.