Taekwondo is one of the most popular martial arts. Several studies on martial arts have been conducted, especially Taekwondo, using sensors and cameras. One of the most vigorous research projects is the use of IMU sensors for motion and camera detection based on image processing. The use of IMU sensors and cameras is essential as both can provide accurate and real-time data on body movements, with IMUs offering details on orientation and acceleration, while image processing-based cameras provide comprehensive visualization and analysis of movements, so the combination of the two can improve accuracy and efficiency in motion detection and analysis. Taekwondo re-research is important because it can provide innovative solutions for more accurate and objective scoring in competitions, as well as contribute to improving the quality of training for athletes and practitioners, by offering efficient, self-directed, and data-driven methods of learning kick techniques, to accelerate the process of skill mastery and improve performance in competition. In this study, testing was carried out on a dataset taken from Taekwondo athletes consisting of five data classes, namely Upper Ap Chagi, Bottom Ap Chagi, Upper Dollyo Chagi, Bottom Dollyo Chagi, and Dwi Chagi, using the frequency domain. The features in the frequency domain used include Total Power (TTP), Median Frequency (MDF), 1st Spectral Moment (SM1), Peak Frequency (PKF), Mean Power (MNP), Mean Frequency (MNF), and others. With several features, testing will be carried out using several machine learning algorithms. From the test results, the highest accuracy is 97.14% using Random Forest. This result has improved considerably with the approach using the proposed frequency domain because similar tests have yet to be carried out on the dataset.