A novel pencak silat punch pattern recognition approach using supervised learning.

Pencak Silat is a martial sport that contains elements of strength and artistic beauty. There are several punching, kicking and parrying movements in this sport, and scoring is also based on the type of movement performed. So far, punching and kicking movements are still determined manually. Currently the use of technology in developing this technology is limited, both at the training stage for athletes and competitions. This research aims to propose a novel pattern recognition approach for pencak silat punch types classification using supervised learning. This research was carried out in several stages, including hardware design, software design, the process of integrating with the Internet of Things (IoT), data collection, data preprocessing, learning process, testing process, and evaluation process. The data used is primary data by analyzing the pattern of punch movements (xy-axis positions) in pencak silat using hardware and software that has been built. This data is divided into learning data and testing data, which are then used in the learning and testing processes. The learning process uses Self Organizing Map (SOM) and K-Means. The results obtained were the first testing process which was carried out using SOM, obtained accuracy value of 89.00% and an AUC value of 0.91. While accuracy value of the testing process using K-Means obtained an accuracy value of 89.833% and an AUC value of 0.92. It is hoped that the contribution of this research can become a decision support system in determining the type of punch in pencak silat. [ABSTRACT FROM AUTHOR]