标题: | 智慧型手势辨识系统设计 Intelligent Hand Gesture Recognition System Design |
作者: | 洪新光 Hung, Iman 陈永平 Chen, Yon-Ping 电控工程研究所 |
关键字: | 手势辨识;类神经网路;手姿态;Gesture recognition;Neural Network;Hand posture |
公开日期: | 2008 |
摘要: | 本論文主要目的是设计智慧型手势辨识系统,此系统是根据人脑所认知手姿态状态来识别不同的手势,其中有九种手势可以被此系统来描述,包含“向左”、“向右”、“左转”、“右转”、“向上”、“向下”、“热机”、“追纵”和 “训练”。手姿态的认知以及手势的识别可以透过类神经网路的学习来处理,首先可利用触发式类神经网路来达成手姿态的认知,再借由手势分类器来完成手势的识别。其中手势分类器可分为前馈式类神经网路和递回式类神经网路两种类型,虽然两者都可以达到很好的手势识别效能,但是仍以递回式的类神经网路为佳。 The main purpose of this thesis is to design the intelligent hand gesture recognition system, which can recognize different hand gestures according to cognitive posture states of human brain. There are nine hand gestures which can be described by this system, including “Turn right”, “Turn left”, “Upward”, “Downward”, “Right around”, “Left around”, “Warming”, “Following” and “Learning”. The cognition of hand posture states and recognition of hand gestures can be learned by neural network. A hand gesture analyzer, composed of a repeated state retriever and a gesture classifier, is applied to recognize the hand gestures. The hand gesture is closely related to the change of hand posture states; therefore, a repeated state retriever is used to turn hand posture state sequence into triggered state sequence, which can be further classified by the gesture classifier. The gesture classifier can be implemented by two types of neural network, feed-forward and recurrent. It can be shown that both types of gesture classifier can well recognize the hand gestures. However, since the feed-forward classifier is often interfered by undefined hand posture state sequence, the recurrent classifier has a better result in had gesture recognition. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#GT079612543 http://hdl.handle.net/11536/41859 |
显示于类别: | Thesis |
文件中的档案:
If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.