Motor Function Evaluation System Based on Machine Vision for COPD Patients

For the home-based rehabilitation of patients with stable chronic obstructive pulmonary disease (COPD), if effective rehabilitation training is carried out without the participation and supervision of doctors, it is an urgent problem to be solved and a hot spot of clinical research. Therefore, rehabilitation training and evaluation method based on ordinary camera motion recognition tracking and machine learning algorithm is proposed. Special normalization is used to solve the position difference caused by the difference in human body shape or the distance from the camera. Dynamic time warping algorithm is used to solve the problem of movement lag caused by the decline in movement ability of COPD patients. Support vector machine (SVM) algorithm is used to establish the evaluation model which is mapped with the evaluation of clinical experts. In order to verify the performance of the system, a clinical trial of Baduanjin Exercise in 18 COPD patients was carried out. The results show that the accuracy of motion evaluation algorithm is 98.4%, and it has a positive linear relationship with the rehabilitation expert score (r = 0.967, P<0.01). The frame rate is about 30fps on the ordinary PC (NVIDIA GTX 1650), which can be used on the home computer with GPU graphics card.