The relationship between human body keypoints and psychological states with machine learning: a pilot study

Measuring an individual’s psychological state from gait videos is an important challenge. Currently, the relationship between a gait video and the psychological state is not clear, it is difficult to measure the psychological state from the video. Therefore, this study aims to investigate the relationship using machine learning. We propose developing the psychological state classification model that utilizes masked time-series modeling and human body keypoints obtained from human pose estimation. The results of two prediction models indicated that some psychological scales are related to the gait video.