Javanese Traditional Dance Movement Evaluation System Based on Pose Estimation with Camera and Arduino for Artist Training

Abstract

Training in Javanese traditional dance requires detailed correction of posture and limb positions, but limited access to instructors can hinder independent learning and increase the risk of repeatedly performing incorrect movements. This study aims to design and evaluate a Javanese traditional dance movement assessment system based on pose estimation using a camera, MediaPipe Pose, OpenCV, and Arduino as an independent training medium for artists. An experimental design method was applied through system requirement identification, hardware and software design, reference dataset preparation, pose detection algorithm implementation, feedback integration, and system accuracy and response testing. The camera captured dancers' movements in real time, while MediaPipe Pose extracted 33 body landmarks. The user's pose coordinates were compared with reference poses using Euclidean distance and a predetermined threshold. The system was tested on five basic Gambyong dance movements involving five participants from beginner to advanced levels. Each participant performed every movement three times, producing 75 movement samples. The results showed an average accuracy of 87% under lighting conditions of 300–500 lux and movement speeds of 60–80 bpm. The system detected angular differences of 5–10 degrees, particularly in the arms and legs, and provided visual feedback indicating correct or incorrect movements. Accuracy decreased to approximately 65–70% under low lighting, fast movements, and partial body occlusion. Therefore, the system is feasible for further development as an interactive and economical dance training tool that supports cultural preservation, although dataset enrichment and improved detection of complex movements remain necessary.

Keywords: Arduino, camera sensor, MediaPipe, traditional Javanese dance, motion detection