Soccer Player Motion Recognition Based on Statistical Weights

Abstract

In this work, we address a new approach to motion recognition of soccer players in a video scene. The inherent motion of a moving character’s arms and legs and neighbouring position of the object, such as a ball, closely influences the motion of the soccer player.

In this paper, we choose the features for recognising a player’s motion by human knowledge and describe correlation between objects. We propose an approach to motion recognition by statistical weights. Also we test, analyse, and estimate the proposed approach. Our algorithm is tried and tested on soccer games. We include successful results as well as a few examples of the algorithm’s failures. Finally, we suggest some improvements and future extensions.