基于Kinect的大屏幕手势互动系统研究与实现

发布时间：2018-03-09 15:47

本文选题：人机交互　切入点：Kinect　出处：《安徽大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着计算机、信息化的发展,人机交互在办公以及生活中显得越来越重要,自然用户界面已经成为人机交互发展的趋势,目前研究比较热门的有用语言、动作及表情等与计算机进行交互。手势识别作为自然人机交互的一项关键的技术,具有灵活、直观、简单等优点,成为动作与计算机交互研究领域中的热点,在业内引起了很多学者的关注和研究。手势识别的发展由早起的佩戴数据手套逐渐的转向基于视觉方法的研究,目前大多的手势识别研究是在计算机视觉的基础上进行的,通过摄像头直接获取背景空间的信息,无需佩戴复杂、昂贵的输入设备。众所周知的触摸屏技术现已发展的比较成熟,但大多应用于小屏幕的操作上,例如:手机、笔记本、小型触摸屏等。针对大屏幕的操作,目前缺乏较有效的交互手段。本文对动态手势识别进行了研究并设计了一套动态手势识别交互系统,实现了人与大屏幕的实时互动操作。Kinect是微软2010年研发出来的,能够在复杂的背景环境下快速准确的获取人体骨骼节点信息。本文以Kinect为输入设备,获取人体手部、手腕、肘部和脊椎这几个关键的骨骼节点信息作为实验的特征向量。主要工作有以下几个方面:针对手势识别的研究,分析了整个识别过程,对识别过程中每个阶段采用的技术做了详细的调查和研究。阐述了 Kinect设备在获取深度信息上的强大功能以及基于Kienct手势识别研究的现状和优点。研究了动态手势的识别过程,针对手势识别起点和终点的检测,本文提出一种基于特殊手势判断的方法。定义一种静态手势,通过识别定义的手势的判断动态手势是否开始,对于动态手势的结束则根据计算连续多帧内特征节点运动的距离判断。在动态手势识别过程中,针对手势识别的效率和速率问题,对传统的动态时间规整(Dynamic Time Warping,DTW)算法进行了改进。首先,利用特征节点在运动过程中的方差对其进行了权值的动态分配,以便提高识别效率。其次,对DTW算法的搜索路径进行了详细的分析,采取点与线结合的方法对其搜索路径进行约束,减少计算量,提高识别速率。最后,将改进后的DTW算法与K近邻(K-nearest neighbor,KNN)算法进行结合用于识别手势,能够有效的提高识别效率。最后,设计并实现了一个与手势识别相关的大屏幕互动系统,在系统中手势代替了鼠标的部分功能,直接通过不同的手势对大屏幕上的内容进行操作,整个系统运行流畅,让人真实的体会到人机交互。
[Abstract]:With the development of computer and information, human-computer interaction is becoming more and more important in office and daily life. The natural user interface has become the trend of human-computer interaction. Gesture recognition, as a key technology of natural human-computer interaction, has the advantages of flexibility, intuition and simplicity, and has become a hot spot in the field of interaction between action and computer. The development of gesture recognition has gradually shifted from the early wearing of data gloves to the research based on visual methods. At present, most of the research on gesture recognition is carried out on the basis of computer vision. Access to background information directly through the camera without the need to wear complex, expensive input devices. The well-known touch screen technology is now relatively mature, but it is mostly used in small screen operations, such as mobile phones, laptops, etc. Small touch screen and so on. For the operation of large screen, there is a lack of effective interactive means. In this paper, dynamic gesture recognition is studied and a set of dynamic gesture recognition interaction system is designed. The real-time interactive operation of human and large screen. Kinect is developed by Microsoft in 2010. It can quickly and accurately obtain human skeleton node information in complex background environment. In this paper, Kinect is used as input device to obtain human hands and wrists. Elbows and vertebrae are the key information of bone nodes as the feature vectors of the experiment. The main work includes the following aspects: according to the research of gesture recognition, the whole recognition process is analyzed. This paper makes a detailed investigation and research on the technology used in each stage of the recognition process, expounds the powerful function of Kinect equipment in obtaining depth information and the present situation and advantages of the research on gesture recognition based on Kienct, and studies the process of dynamic gesture recognition. Aiming at the detection of the beginning and end of gesture recognition, this paper proposes a method based on special gesture judgment, which defines a static gesture and determines whether the dynamic gesture begins or not by recognizing the defined gesture. For the end of dynamic gesture, the distance of motion of feature nodes in successive frames is calculated. In the process of dynamic gesture recognition, the efficiency and rate of gesture recognition are discussed. The traditional dynamic Time warping (DTW) algorithm is improved. Firstly, the weights are dynamically assigned by the variance of feature nodes in the process of motion in order to improve the recognition efficiency. The search path of DTW algorithm is analyzed in detail, and the search path is constrained by the combination of point and line, which can reduce the amount of calculation and improve the recognition rate. The improved DTW algorithm and K-nearest neighbor KNN algorithm are combined to recognize gestures, which can effectively improve the recognition efficiency. Finally, a large screen interactive system related to gesture recognition is designed and implemented. In the system hand gestures replace some of the functions of the mouse directly through different gestures on the large screen content to operate the whole system runs smoothly let people truly understand the human-computer interaction.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】