Tai Chi Exercise Posture Detection and Assessment for the Elderly Using BPNN and 2 Kinect Cameras

Abstract

Exercise and recreation are beneficial to all genders and ages, exercise reduces stress and makes people healthy. Physical limitation among the elderly is the major concern and needed to be taking care for the elderly exercise. Low-impact exercises such as walking, slow jogging in the park, and Tai Chi are recommended for the elderly. Tai Chi is a slow and gentle exercise, which can help the circulatory system and dementia in the elderly; it also helps the elderly to get socialized and make new friends. Unfortunately, in the COVID-19 pandemic, people must stay in the house and avoid social activities including outdoor exercises and recreation. This paper aims to develop Tai Chi exercise posture detection and assessment system for helping the elderly to practice Tai Chi at home by themselves. The system provides Tai Chi video clips for demonstration and the graphics user interface (GUI) for capturing the movement of the elderly while they are exercising Tai Chi. The system will detect and assess the elderly's movement whether it is correct or not by using 2 Kinect cameras. The Kinect is used for joints detection and the series of joints movement will be used to compare with the correct Tai Chi postures stored in the system. The questionnaire, which was developed based on the usability criteria defined by the ISO 9241–11 and the users' experience, was used to evaluate the system. The precision, recall, F1-score, and accuracy of our system are 0.94, 0.98, 0.96, and 0.93 respectively.