Athlete Action Recognition in Sports Video: A Survey

Abstract

The world of sports is growing in rapid and precise shifts which are more demanding for coaches and trainers to track an athlete’s actions. Athlete action recognition is now trending as an ordinary and necessary practice in sports video analysis. The Goals of the sports team are to identify the weakness of the opposite team and to improve its own-potential strength. In such cases, athlete action recognition of a particular player helps in increasing the player’s performance, enhances coach training, and reduces injury risk. It provides a rich set of opportunities for athlete tracking which helps the trainers, coaches and competitors. The existing studies of the athlete action recognition based on team sports videos and where the particular player’s identity issues lie are not adequate. To overcome the existing issues this paper focuses on reviews about athlete action recognition, on particular athlete tracking, and also on the sports video datasets. The paper focuses on some of the common issues like frequent occlusion, background clutter, out of view, motion blur and Inter-class similarity. In this paper the review parts are divided into two categories: Athlete tracking and sports video action recognition, where the athlete tracking parts studies the player’s identity and the sports video action recognition part reviews the action performed by the player, using learning algorithms. Finally, the future direction of the athlete action recognition on particular athlete tracking in sports video is recommended based on current technologies.