Evaluating Simple Exercises with a Fuzzy System Based on Human Skeleton Poses
Chyan Zheng Siow, Wei Hong Chin, Naoyuki Kubota · 2023
Exercise is important for elderly people to maintain their physical ability. However, evaluating the quality of exercise requires human judgment, which is subjective and time-consuming. In this paper, we propose a method that uses a fuzzy system to evaluate a simple exercise (overhead arm clap) based on human skeleton poses extracted from video sequences. Exercise is important for elderly people to maintain their physical ability. However, evaluating the quality of exercise requires human judgment, which is subjective and time-consuming. In this paper, we propose a method that uses a fuzzy system to evaluate a simple exercise (overhead arm clap) based on human skeleton poses extracted from video sequences. Our method consists of four steps: (1) converting the video into a sequence of human skeleton poses using the MediaPipe tool, (2) extracting the essential poses from the sequence using a fixed incremental adaptive resonance theory, (3) generating the fuzzy membership functions based on the pose similarity and the completeness of the exercise, and (4) performing rule inferencing and defuzzification to obtain an exercise score. We evaluate our method on four types of videos: (1) correct exercise, (2) incorrect pose, (3) incomplete sequence, and (4) different exercise (squat). Our results show that our method can distinguish these different types of videos and provide a reasonable exercise score. We conclude that our method is a promising technique for evaluating simple exercises using human skeleton poses.