A Virtual Training System for Skill Improvement of Well–timed Attack in Karate Match

Abstract

Martial arts experts have excellent attack skill that takes advantage of a momentary chance. Mastering the skill requires a long years training as repeated sparring, while such training puts players in great danger of injuring either themselves or sparring partners. In this paper, we describe a virtual training system that provides a player with a virtual sparring partner for skill improvement of well–timed attack in Karate match. It is required that the virtual partner plays the role of defense having a knowledge of the skill. Unfortunately, little is known about the knowledge because the knowledge is tacit. Therefore, we propose a method which extracts the knowledge from motion data sets of Karate matches games using a data mining method and determines the defensive property of the virtual partner based on the extracted knowledge.