Motor Skill Development Using Motion Recognition based on an HMM
Keita Yamada, Kenji Matsuura, Keisuke Hamagami, Hirofumi Inui · Procedia Computer Science · 2013
The purpose of this study is to propose an original method that enables the recognition of a swinging motion. Furthermore, we apply the method for supporting skill development. In this study, we use an optical-motion capture system to monitor a human's motion. The captured data are transformed into series of symbols by using a feature quantity. The feature quantity is height of hands. The symbol sequence is learned and recognized by a Hidden Markov Model (hereinafter HMM). An evaluation was performed with five subjects. The target motions are the batting and pitching of a baseball and, the motor actions of tennis, such as swinging in service, forehand and backhand. The accuracy of evaluation with the dataset is more than 80% recognition.