Developing the evaluation system of the Thai dance training tool

Abstract

© 2019 IEEE. The Thai traditional dance training tool was created with the objective to develop the technologies for supporting the learning of Thai traditional dance, which still has been transferred orally from generation to generation. The tool collects the dataset of dance movements from traditional Thai dance experts using motion capture system and it uses them to show a demonstration for the user, then it gives the user a stage to perform and record their movement data using the Microsoft Kinect sensor. The movement data obtained will be evaluated and a score is generated to provide feedback for the user at the end. The project also tests the performance of the evaluation by comparing the score generated by the tool with the score generated by a traditional Thai dance teacher from the College of Dramatic Arts to calculate the inverted pair and the displacement in order to see the difference. The project tries to test the evaluation method that uses the formula to calculate the similarity percentage between the user's movement and the expert's movement. This testing method gives 17.7% inverted pair and 53% displacement. The project and also tries many evaluation methods using different regression algorithms in machine learning from which the best results are from the Random Forest Classifier that gives 7.89% inverted pair and 48% displacement, and from the Decision Tree Regressor that gives 15.79% inverted pair and 44% displacement.