Context aware exercise recognition with body area network for rehabilitation

Abstract

Rehabilitation exercise is important for patients to regain ability to perform daily activities. However, the challenge with such rehabilitation is the limitation of resources as well as inconvenience for patient, which leads to the rise of in-home rehabilitation exercise. The common approach is to use miniature wearable sensors to monitor human motions and movements in rehabilitation exercise. However, change in sensor contexts results in incorrect or imprecise motion recognition. The context considered is the orientation of sensor with respect to the monitored body segment. The motion recognition algorithms must be able to recognize the context change and adapt automatically to provide correct motion recognition without constraints on assumption of unchanged contexts. As a preliminary, this project will study the issues faced with non-context-aware motion recognition methodologies in measuring joint angle in arm ROM (Range of Motion) rehabilitation exercises using Inertial Measurement Unit (IMU). Several methods were proposed to obtain joint angle: DCM Method, DH Kinematic Model, and Joint Coordinate System. From all the three methods, it was observed that the measured joint angle is affected by sensor context change. Future research is needed to quantify the context change and provide the necessary compensation to obtain correct joint angle measurement in the event of context change.