Utilization of Pose Estimation and Multilayer Perceptron Methods in the Development of Taekwondo Martial Arts Independent Learning

Abstract

Taekwondo is a martial art from South Korea that has been developing in Indonesia since 1975 in North Jakarta. Since then, Taekwondo has become increasingly popular, and it can be seen when Taekwondo entered the official sport in the XI PON arena in 1985. Due to the popularity of Taekwondo, many instructors have built Taekwondo learning clubs throughout Indonesia. Not only that, the championship for Taekwondo has also increased rapidly in Indonesia. Due to many tournaments that are held, many Taekwondo clubs carry out intensive training to train young athletes to participate in the tournament. Still, the training is considered less than optimal due to the large number of students participating in the training, which makes the instructor pay less attention. To solve the problem, the author has an idea to build an independent learning system using the Pose Estimation method, which is used so that the computer can recognize Taekwondo movements and Multilayer Perceptron with Backpropagation learning which is used to predict Taekwondo movements, By utilizing Pose Estimation and Multilayer Perceptron, machine learning models can be built that can predict Taekwondo movements in real-time which can help Taekwondo students to learn independently from home. This study uses primary data obtained from the DAS (Dynamic Able Success) club, containing two kicks and two blocks. After conducting and evaluating a series of experiments, this study got the most optimal accuracy of 100%.