Adaptive Sensor Fusion And Machine Learning Framework For Movement Detection In Rehabilitation Systems

This paper presents an adaptive sensor fusion and machine learning framework for movement detection in rehabilitation systems. Accurate movement monitoring is crucial for assessing patient progress during rehabilitation. Traditional manual methods are often subjective and time-consuming, limiting their effectiveness. The proposed system integrates data from wearable inertial measurement units (IMUs) and machine learning models to classify movement performance into three categories: Optimal, Slight Deviation, and Deviation. The framework utilizes sensor data, including joint kinematics, stride parameters, and energy-related features, to analyze patient movements. Machine learning models, such as support vector machines, decision trees, random forests, and gradient boosting, are employed to classify movement patterns. The feature selection process ensures that only the most relevant features are used, improving model accuracy and efficiency. The system is implemented as a web-based application, offering a scalable and efficient solution for rehabilitation, providing automated feedback to clinicians and patients, and enhancing the rehabilitation process.

Keywords: Sensor Fusion, Rehabilitation, Movement Detection, Machine Learning, Wearable Sensors, Feature Selection, Classification, Inertial Data, Adaptive Framework, Feedback.