HAND TRACKING AND GESTURE RECOGNITION FOR PLAYING VIRTUAL GAMELAN

• Main Authors: Wisnu Aditya, 何迪亞
• Other Authors: Timothy K. Shih
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/hy7h37

碩士 === 國立中央大學 === 資訊工程學系在職專班 === 105 === People who live in modern society usually forget their culture by degrees, they prefer about modern thing rather than traditional. It cannot be left continuously in this kind of situation, because if it continuous someday traditional culture will vanish, so we need to preserve our traditional culture a creative and innovative way. One kind of the traditional performing arts from Indonesia is Gamelan since the 6th century. Combining the traditional culture and modern technologies is expected to solve this problem. This combination is implemented in the form of a virtual gamelan system. The system is controlled in real-time using hand gestures to make it look like the original gamelan play. Using gesture for playing gamelan provide a new experience to the players of gamelan. The use of a hand gesture offering an optional intelligent and natural way to interface tools for human computer communication. Hand segmentation and tracking are the biggest issues in any hand-gesture recognition application and they provide the most vital input for the succeeding gesture recognition algorithm. Using depth data can speed up the process of segmentation because the depth data has information that can recognize the position of an object, then we can separate objects and backgrounds easily. We do segmentation using the threshold method, this threshold will be able to reduce the amount of data to be processed so as to speed up the computation process. In this research, we propose Density-based spatial clustering of applications with noise (DBSCAN) for a data clustering algorithm. This method used in both hand tracking and hand gesture recognition. Hand tracking process uses DBSCAN to obtain hand classes, DBSCAN is expected to produce two classes representing the right hand and left hand. However, these two classes need to be labeled so that no class changes in the next frame. Other gestures using distance measurement methods. The distance is obtained from the position of the hand center between the classes in the current frame with the previous frame. Finally, we did some experiments to get the best parameters for DBSCAN, this parameter will produce the best result. Then we tested the system by playing in various poses. The average accuracy of hand gesture that using DBSCAN is 92%. The results show that our method performs well on a virtual gamelan system.