【作者】 龚陶波；

【导师】 李志扬；

【作者基本信息】 华中师范大学 ， 物理电子学， 2008， 硕士

【摘要】 摄像机能够捕捉到用户通过内在自然方式(如眼神、表情、手势和动作等)所表达的信息,因此通过摄像机与计算机进行交互是一种高效自然的人机交互方式,它能使人机对话变得简单,减少人们使用计算机的障碍。特别是最近几年,随着计算机技术的迅猛发展,研究符合人机交流习惯的新颖人机交互技术变得异常活跃,也取得了可喜的进步。这些研究包括人脸识别、面部表情识别、唇读、头部运动跟踪、凝视跟踪、手势识别、以及体势识别等。手势是一种自然、直观、易于学习的人机交互手段。手势识别按输入设备不同可以分为基于数据手套的手势识别和基于计算机视觉的手势识别。其中基于计算机视觉的手势识别以人手直接作为计算机的输入设备,人机之间的通讯将不再需要中间媒体,用户可以简单地定义一种适当的手势来对周围的机器进行控制。但是由于手势本身具有多样性、多义性以及时间和空间上的差异性等特点,加之人手是复杂变形体以及视觉本身的不适定性,基于视觉的手势识别是一个富有挑战性的、多学科交叉的研究课题。本文设计实现了一个基于计算机视觉的静态手势识别系统,该系统能够实时地对从摄像头输入的10个常用静态手势进行识别。系统的设计准则一是实时性,二是准确性。在手势建模方面,采用基于表观的手势模型;在手势分析方面,经过手势图像预处理和特征参数提取得到八个手势特征参数;在手势识别方面,采用二次分类(粗分类和细分类)的方法进行识别。整个系统分三个部分实现。手势图像预处理部分,根据人体的肤色特征从环境中分割出手区域,然后通过图像增强和拉普拉斯边缘提取算法得到手势轮廓;手势特征提取部分,提取了八个手势特征参数,组成特征向量;视频流实时处理部分,使用天敏SDK-2000图像视频采集卡,通过回调函数对摄像头输入的视频流进行计算,提取出单个静态手势图像,并进行实时地识别。更多还原

【Abstract】 Camera can catch the information naturally expressed by people such as eye movements, expression, gesture and motion. For users, communicatingwith computers by camera is efficient and natural. It can provide a convenient addition to user-computer dialogues and reduce the obstacle to use computer. Recent years espacially, with the development of computer science, research on new human-computer interaction technology become extremely active, and advancement has been achieved. These research includes face recognition, expression recognition, hand gesture regognition, pose recognition and so on.Hand gesture is a natural and straight human-computer interactionmethod. There are two methods on hand gesture recognition, recognition based on data glove and recogniton based on computer vision. Take hand as the input equipment directly, communication between human and computer will need no more other intermediate media. Users can control the machines around simply sign to it with the hand gesture user itself defines. However, gesture has the characters of multi-mode, multi-meaning and has discrepancy undercertain time and space situation; moreover, human handsare complicated transformed objects and there is visual instability, all of which make gesture recognition based on computer sight become a challengeable multi-subject researchgoal.This paper realized a static hand gesture recognition system based on computer vision. This system can recognize 10 common static hand gestures inputted from camera at real time. It is a real-time system, so both recognition time and the correctrecognition rate have to be considerd while designing the system. In the aspect of hand gesture modeling, the system adopt hand gesture model basedon apparent; In the aspect of hand gesture analysis, the system picks up eightcharacters through image preprocess and character extraction; In the aspect of recognition, the system adopts two times classifation (rough classification and particular classification).The system consists of three part. First, preprocession of the originalhand gesture image, in this part hand area is extracted from background through the character of human complexion, and then the system gets the edge through noise smoothing and Laplacian edge extraction; Second, extraxction of hand gesture characters, in this part the system extracts eight characters as an eigenvector; Third, the real-time procession to the Video Capture Card (天敏SDK-2000) , in this part the system processes the real-time data inputted from camera with a callback function and find the single image of static hand gesture.更多还原