Gesture spotting algorithm for door opening using single wearable sensor

Rehabilitation programs are critical for stroke patients to regain daily living skills. Accurate and objective assessment of the motor function is essential for health professionals to determine the appropriate clinical interventions and rehabilitation programs. Traditional assessment methods based on questionary and clinical observation have some issues in episodic examinations, subjective assessment and limited medical resources which tend to cause the difficulties for clinical professionals to observe real performance of daily activities performed by the patients in the home environment. In this paper, a door-opening activity monitoring system using a single wrist-worn inertial sensor to support assessing the performance of upper-limb movement during daily living is developed. The proposed approach could spot and identify three sub-gestures from complete door-opening gestures, including grasping door handle, rotating handle, opening door. The overall performance of the proposed approach achieves 0.933 in accuracy, 0.896 in precision and 0.904 in recall.