A Real-Time Hand Tracking and Gesture Recognition System Based on Skin Color Region Locating

• Main Authors: Kai-Lun Chang, 張凱倫
• Other Authors: Chang-hong Lin
• Format: Others
• Language: en_US
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/68084913421277017380

碩士 === 國立臺灣科技大學 === 電子工程系 === 105 === Nowadays, intuitive operation modes with the use of gestures have become more and more popular in many applications, such as human-computer interaction, surveillance, and gaming. Because of its convenient and intuitive manipulation, traditional remote and input device control are going to be replaced with hand-gesture based controllers gradually. In this thesis, we propose a real-time hand tracking and gesture recognition system based on skin color region location, and the recognition process comprises the face detection, hand detection, hand tracking, and gesture recognition. For better accuracy, we take the location of detected skin color regions to correct the tracking results from traditional CamShift algorithm. Four dynamic hand gestures consisting of leftward, rightward, upward and downward are able to be recognized. In the experiment, the performance of the proposed system was verified for both right and left hands for ten different users. The experiment results show that our proposed approach is effective, with highly increased recognition accuracy, without wasting too many computations, and can avoid the fail to track problem under the situation that some areas in the background have the similar color distribution as the target. The techniques developed in this study can not only be used in human-computer interaction, but also be applied to augmented reality application to make the user experience much better than before.