Abstract:
This research aims to automatically check the correctness of knee exercises for patients with knee osteoarthritis. The patients need to be treated by physical exercise for knee but cannot know the efficacy of physical exercise while doing at home. We developed the automatic system which processes images from knee exercising video of subject who does exercise. This system utilized of histogram analysis and skeletonization. The experimental result was tested by 17 subjects. The system verifies whether subjects do an exercise correctly or incorrectly. The overall accuracy of the exercise postures classification is 73.94%, the overall accuracy of the exercise postures correctness checking is 90.18, and the overall accuracy of the automatic system which verifies the exercise postures is 88.12%. The advantage of the purpose is the patient can do physical exercise spontaneously without attaching a device to the body and know their performance at home. © 2020 IEEE.