A Virtual Training System for Supporting Tactical Skill Improvement in Martial Arts Match

支持武术比赛战术技巧提升的虚拟训练系统

• 批准号: 18500744
• 负责人: TANAKA Kazumoto
• 金额: $ 1.63万
• 依托单位: Kinki University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 2006
• 资助国家: 日本
• 起止时间: 2006 至 2007
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-18500744/
• 关键词: Martial Arts; Match Tactics; Open Skill; Physical Skill; Sequential Pattern Mining; Motion Sequence; Motion Capturing; Virtual Space; 対人競技; 教育工学; 画像認識; 人工知能; バーチャルリアリティ; データマイニング; 動作マイニング; 熟練技

(1) Athlete's tactics extraction methodThe method for extracting the feature motion pattern in the offense and defense process of the martial arts match was developed by using sequential pattern mining. First of all, the offense and defense process is expressed by the tuple of a motion label and a positional label in the motion section. Next, the method extracts motion sequences that have a positive correlation with a successful attack and a negative correlation with a failure attack from the database of the offense and defense process. When we analyzed the attack pattern of two Karate players who had National-Sports-Festival (Koku-Tai) victory experience by using the method, each player's feature attack pattern was able to be extracted and was found to coincide with the opinion of a Karate specialist. Moreover, it was clarified that the successful attacks have a strong correlation with the state immediately before each attack according to extracted patterns from the various athlete's matches (130 matches). This result agrees with teaching in martial arts from old.(2) Virtual training system for skill improvement of well-timed attackWe have developed a virtual training system for mastering the timing of successful attack based on the result of the tactics pattern extraction. First of all, we designed state transition rule to move a CG player (opponent player) who could give the timing of successful attack to real player (a trainee). Next, the mechanism that the system evaluates whether a trainee correctly reacted to CG player's state transition was clarified. This became possible because captured trainee's motion and CG player's motion could be compared with the tactics pattern. Six student athletes evaluated the system, and the high appraisal was obtained as an independent training system accordingly.

(1)运动员战术模式提取方法利用序列模式挖掘技术，提出了武术比赛攻防过程中特征运动模式的提取方法。首先，进攻和防守过程由运动部分中的运动标签和位置标签的元组表示。接下来，该方法从进攻和防御过程的数据库中提取与成功攻击具有正相关性并且与失败攻击具有负相关性的运动序列。当我们分析的攻击模式的两个空手道选手的国家体育节（国大）胜利的经验，使用该方法，每个球员的特征攻击模式能够被提取出来，并发现符合空手道专家的意见。此外，根据从各种运动员的比赛（130场比赛）中提取的模式，明确了成功的攻击与每次攻击之前的状态具有强相关性。这一结果与自古以来的武术教学相吻合。(2)基于战术模式提取结果的攻击时机掌握虚拟训练系统。首先，我们设计了状态转换规则，将一个能够给出攻击成功时机的CG玩家（对手玩家）转移给真实的玩家（练习生）。其次，阐明了系统评估受训者是否正确地对CG玩家的状态转换做出反应的机制。这成为可能，因为捕捉到的学员的运动和CG球员的运动可以与战术模式进行比较。6名学生运动员对该系统进行了评价，作为一个独立的训练系统得到了较高的评价。

项目成果

空手道競技の戦術修得のためのデータマイニング

用于学习空手道比赛战术的数据挖掘

• 发表时间: 2008
• 期刊: 人工知能学会研究会資料SIG-ALST A703
• 作者: 長谷川元洋(編著);他;田中 一基;田中 一基
• 通讯作者: 田中 一基

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

空手道比赛适时进攻技术提高虚拟训练系统

• 期刊: Transactions of Japanese Society for Information and Systems in Education Vol.24, No.4(now printing)
• 作者: 田中一基;長谷川誠;黒瀬能聿;小林規矩男;Kazumoto Tanaka
• 通讯作者: Kazumoto Tanaka

A Data Mining Method for Learning Tactics in Karate Match

空手道比赛战术学习的数据挖掘方法

• 发表时间: 2008
• 期刊: The Japanese Society for Artificial Intelligence SIG-ALST-A703
• 作者: 長谷川元洋(編著);他;田中 一基;田中 一基;Kazumoto Tanaka
• 通讯作者: Kazumoto Tanaka

A Motion Mining Method for Supporting Skill Improvement in Martial Arts

支持武术技能提高的运动挖掘方法

• 发表时间: 2007
• 期刊: IEICE Technical Report Vol.106, No.507
• 作者: 田中 一基;ほか3名;田中 一基;Kazumoto Tanaka
• 通讯作者: Kazumoto Tanaka

武道の修得支援のための攻防スキル抽出法

攻防技能提取方法，支持武术习得

• 发表时间: 2007
• 作者: 田中一基;長谷川誠;黒瀬能聿;小林規矩男;Kazumoto Tanaka;田中 一基
• 通讯作者: 田中 一基