A Generic Framework for Sport-specific Movement Recognition

摘要

In this paper, we propose a generic framework for sport-specific movement recognition. The framework is based on the technology of the wireless wearable sensing device (WSD) which was adopted to distinguish the movements of table tennis athletes. The system has the advantages of low price, convenient use and high calculation efficiency compared to traditional video approaches. In order to validate our system, ten table tennis players put the WSD on their left/right wrists and carried out different table tennis movements in a real table tennis court. We use five different classification methods to identify different motions of table tennis from the recorded data. The experiments show that our generic framework can distinguish the movements of table tennis i.e., sidespin serves, backhand chops, backhand pulls, backhand pushes, flat serves, underspin serves, with an average accuracy of 94%. The main purpose of developing this generic framework is to change the manner from athlete-level evaluation based on subjective experience to athlete-level evaluation based on objective data.