Evaluating Karate-do movements using Kinect V2 camera and tools.

Abstract

Karate-do is a martial art of Japanese origin that has a wide variety of techniques. In Karate-do, practitioners must constantly train to perfect their movements. This can be accomplished during a karate class with the help of a teacher or through self-taught practice, done through consultation and repetition of the movements shown in books or instructional videos. However, there is some difficulty for the practitioner to define the efficiency of his/her movements in self-taught practice. In this work, we investigate the effectiveness of using a 3D capture device and machine learning algorithms aiming to detect and automatically evaluate and propose corrections to the movements of karate learners.