A Comprehensive Study on Intelligent Glove-Based Hand Gesture Recognition

• Main Authors: Kuo-Jung Wang, 王國榮
• Other Authors: Chin-Shyurng Fahn
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/47748886670358044214

碩士 === 國立臺灣科技大學 === 電機工程系 === 89 === Along with the continuous changes and improvements of computer systems, data gloves are widely used in virtual reality, sign language systems, man-machine interfaces, and so on. Then the research of gesture recognition becomes more and more important. In general, hand gestures can be grouped into static and dynamic ones. Because dynamic gestures contain both space and time signals, we use the dynamic time warping and frame methods to take the characteristics of the gestures. In our experiments, such two methods perform similar recognition results. Compared to dynamic time warping, the frame method is easier to implement, which is appropriately applied to the real-time systems. In this thesis, we propose a new hand gesture recognition method based on a distance gray relation network. It combines the structure of a neural network with the distance gray relation method to make the latter have the ability of learning. Experiments reveal that the convergence speed of the distance gray relation network is very fast and the recognition rate is very high. In addition, we investigate the gray relation method, back-propagation network, radial basis function network, and hidden Markov model to recognize hand gestures. And we study the effect of distinguish coefficients on the performance of the gray relation method as well as the effect of different learning rates on the performance of the neural networks. Finally, we conclude both advantages and disadvantages of the aforementioned methods. In the future, researchers can choose an appropriate recognition method according to different gesture characteristics, numbers of classes, and numbers of patterns. Therefore, the development of a high robust and adaptive gesture application system can be accomplished.