Classification of Scissor Kick in Taekwondo using Wearable Motion Capture and Machine Learning

Taekwondo is a martial art that aims to defeat the opponent athlete with superior foot and hand techniques. The judges determine scoring according to the type of action performed by the athlete and its effect. In addition, the coach continually observes the athletes throughout training, and feedback serves the athlete’s improvement. The judges or coaches, on the other hand, may make subjective assessments in specific situations. To minimize this, electronic and machine learning-based applications have recently been recommended. Therefore, first and foremost, taekwondo moves must be classified successfully. Scissor kick in taekwondo is a type of kick that requires an advanced level and is quite difficult. Athletes can gather data on how effectively they perform by using motion capture devices, which enables them to assess their performances objectively. Thus, in this study, taekwondo scissor kick movements were collected by a total of 11 athletes, 6 black belts and 5 blue belts with red stripe belts, using a wearable motion capture system. For this movement, temporal features were extracted by calculating the range of motion (RoM) in three axes of beginner and professional athletes’ hips, knees, and ankles. In the classification performed using SVM, kNN, EL, and logistic regression (LR) models, the best performance was obtained with SVM and LR with 90.9%. This study offers an objective feedback system for athletes performing the correctly classified scissor kick movement.