Archery shots visualization by clustering and comparing from angular velocities of bows

Abstract

In individual competitions consisting of repetitive movement sports, it is necessary to increase the reproducibility of movements by recognizing and correcting movement changes per second. Since it is difficult to obtain sufficient awareness only by subjectivity, a mechanism that can objectively confirm the movement is required. In this paper, we propose a system that can easily search for differences in multiple trial motions by the same person for archery movements. The proposed system uses Dynamic Time Warping to determine the similarity of multiple shots of one competitor from the time-series data from the angular velocity sensor attached to the competitor's bow. Based on the similarity distance, K-means Clustering is performed. In addition, the video corresponding to the time at which there is a difference is cut out from the video recorded simultaneously to the sensor data, and the two images are superimposed and presented to visualize the difference. When the proposed system was tested with five intermediate- and advanced-level archers, it was possible to detect differences such as minor shaking, the posture, and the motion speed for approximately 0.5 seconds. These differences can be found by advanced-level archers by carefully comparing the videos for many times, but are difficult to identify by intermediate-level archers.Feedback from interviews with the instructor suggested that the differences detected were meaningful to find out the points for improve archery skill.