Evaluation of Machine Learning Models for Movement Classification in Exergame-Based Rehabilitation

This paper explores the innovative integration of exergames and machine learning techniques to enhance rehabilitation outcomes for patients with motor impairments, particularly those resulting from stroke or other neurological conditions. Utilizing the Remote Monitoring Validation Engineering System (ReMoVES) and the Microsoft Kinect sensor, the study captures and analyzes patient movements during Sit-to-Stand exercises—key activities for improving strength, endurance, and functional mobility. The primary aim is to classify these movements as either normal or abnormal, providing insights into recovery progress. Comprehensive motion data, including joint angles, movement trajectories, and timing metrics, are collected and rigorously processed to extract features that differentiate between normal and impaired patterns. Various machine learning algorithms, such as Support Vector Machines (SVM), Extreme Gradient Boosting (XGBClassifier), and K-Nearest Neighbors (K-NN), are employed to assess and classify patient performance in real-time. The system offers immediate feedback, including performance scores, visual cues, and tailored suggestions, which are crucial for maintaining engagement and motivation. Additionally, ReMoVES allows healthcare professionals to remotely monitor patient progress, enabling timely adjustments to rehabilitation plans. Findings show that combining exergames with advanced machine learning significantly improves movement classification accuracy and enhances patient engagement and motivation, providing a promising solution for optimizing rehabilitation processes and improving patient outcomes in both clinical and home settings.