Knowledge Base Model for Choking Detection in College Tennis as a Component of an Expert System

Abstract

Expert system is a tool to transfer human expertise into specific domains by simulating human intelligence, behavior, and thinking. Choking under pressure in sports is characterized as suboptimal performance in certain actions. Expert systems have found their applications in various domains including sports and athletics. However, the concept of choking under pressure is yet to be tackled by the utilization of the expert system model. Therefore, this paper proposes an approach for choking detection in college tennis by designing an essential component of an expert system, knowledge base. The knowledge base model is formed based upon the expertise of tennis coaches with American college coaching experience. The model is designed for three detection levels: general, semi-personalized, and personalized. The detection of choking moments is done from racquet swing speed measurements collected by a swing speed sensor. Study data were collected with eight participants completing a total of 32 tennis sessions. Results suggest practicality of the proposed model and feasibility of choking detection in college tennis from swing speed measurements by showing that choking detection can reach 68% accuracy while keeping the error rate at 15%. The results reveal individual variations of choking patterns among college tennis players as individually oriented detection techniques return superior detection performance. Future plans for the research are to collect measurements with additional participants and to combine the gathered data with physiological parameter measurements such as heart rate and with other sport specific information such as footwork performance.