Analysis of Soccer Actions using Wireless Accelerometers

Abstract

The objective of this research is to develop a system for soccer gait analysis. The system consists of two subsystems; soccer gait pattern classification and soccer gait pattern recognition. This article describes the key features and phases involved in these two subsystems. During the pattern classification, knowledge base is formed using gait analysis system already developed. Elman neural networks (ENN) is used to perform the pattern classification of a soccer gait. Pattern recognition is based on real time soccer play which in turn recognizes soccer gait. The system developed is also responsible to enrich the knowledge base by the inclusion of newly performed soccer gait patterns and by the rejection of gait patterns not any more performed. The prototype developed is capable of recognizing movement patterns of a soccer player, which were not performed during previous training sessions. Although the pattern classification was carried with discreet movement patterns, the prototype was able to recognize these when they occurred in combination, as would happen in a real soccer game.