Nowadays, the judging of boxing competition is mostly based on human interpretation of the frequency of each kind of punches. But there is no study discuss about the test-retest reliability of human judging. The aim of this study was to conduct the test-retest reliability on human judging. We included four classes A referees to judge the male Olympic Boxing Games and female World Cup final competition. A self-develop user interface programwas used to record the success and failure times six kinds of attack punch: left / right jab punch, left / right hook punch, and left / right uppercut punch, and six kinds of defend pose: deflect, block, parry, footwork, sway, and counter punch. We experiment every film three times a day for totally two days. The two-way ANOVA was used to analyzed within subject (within subject R.), between subject (between subject R.), and inter-group correlation coefficient (inter-group R.). The results showed that the within subject R. of attack punch was 0.73-0.99, defend pose was 0.28- 0.99. The inter-group R. of attack punch was 0.67-0.76, defend pose was 0.08-0.69. The conclusion was that human judging on boxing competition film was based on single view and experience of the referee, lack of presence on site was the limitation, reducing the test-retest reliability of the judging. The combination of different view of the film and three-dimensional motion capture system might solve the problem.
© Copyright 2016 14th International Symposium on 3D Analysis of Human Movement. Published by National Taiwan University. All rights reserved.