基于中文语音交互的虚拟装配技术研究

发布时间：2018-07-11 10:16

本文选题：虚拟装配 + 人机交互　；参考：《北京交通大学》2014年硕士论文

【摘要】：目前,虚拟装配系统通常依赖力反馈、体感设备作为交互工具,采用“抓取”、“拖拽”等直观的手势完成产品的装配。然而,当遇到精细零件模型的装配时,往往达不到预想的结果。针对这一问题,本论文提出了以Kinect体感设备为输入,利用中文语音指令识别引导装配流程,并设计实现了一套基于中文语音交互的自行车虚拟装配系统,验证了方法的可行性。论文的具体研究内容包括： (1)由于基于中文语音交互的虚拟装配涉及到虚拟环境的建模、虚拟环境的渲染、中文语音的识别等多项技术,因此在虚拟装配系统架构设计时重点考虑了多种软件的通讯方式,并以自行车虚拟装配为例设计实现了原型系统。 (2)虚拟装配环境的设计与仿真。利用3DS MAX完成虚拟装配系统中零件和场景的建模,利用Photoshop CS6完成交互菜单设计和材质纹理图片处理,并在3Dvia Studio Pro中导入场景模型,完成了整个虚拟装配环境的设计、场景渲染、装配序列规划等。 (3)为了降低用户的操作错误率,提出了一种基于推荐的零件装配序列规划。通过分析零件的物理连接关系、层次关系以及同层次下装配约束关系,定义了自行车的装配优先级模型。仅当较高优先级的零件完成了装配仿真,其下一优先级的零件才会被系统推荐为待装配零件。 (4)为了解决多人环境下操作者确认问题,提出了一种基于Kinect的中文语音指令识别的方法,并通过实验测试了该方法的识别准确度。该方法借助Kinect设备捕捉多人的骨骼跟踪数据,并据此确定Kinect视野前方一定范围内,距离Kinect最近用户的骨骼信息；然后,利用该用户头部的坐标信息以及Kinect线性麦克风阵列的声源定位功能确定操作者,实时捕获该用户的语音指令并进行识别,通过模拟键盘事件来完成操作用户和虚拟装配系统的交互。
[Abstract]:At present, virtual assembly systems usually rely on force feedback. As an interactive tool, the virtual assembly system uses intuitive gestures such as "grab" and "drag and drop" to complete the assembly of the product. However, when the assembly of fine parts model is encountered, the expected results are often not achieved. In order to solve this problem, this paper proposes to use Kinect as input to guide assembly process with Chinese speech instruction recognition, and designs and implements a bicycle virtual assembly system based on Chinese voice interaction, which verifies the feasibility of the method. The main contents of this paper are as follows: (1) Virtual assembly based on Chinese speech interaction involves modeling of virtual environment, rendering of virtual environment, recognition of Chinese speech, and so on. Therefore, in the design of virtual assembly system architecture, the communication mode of many kinds of software is considered, and the prototype system is designed and implemented by taking the bicycle virtual assembly as an example. (2) the design and simulation of virtual assembly environment. Using 3DS Max to complete the modeling of parts and scenes in the virtual assembly system, using Photoshop CS6 to complete the interactive menu design and texture image processing, and importing the scene model in 3Dvia Studio Pro to complete the design and rendering of the whole virtual assembly environment. (3) in order to reduce the operating error rate of users, a part assembly sequence planning based on recommendation is proposed. The assembly priority model of bicycle is defined by analyzing the physical connection relation, hierarchical relation and assembly constraint relation of the same level. Only when the higher priority part has completed the assembly simulation, the next priority part will be recommended as the part to be assembled. (4) in order to solve the problem of operator confirmation in multi-person environment, A method of Chinese speech instruction recognition based on Kinect is proposed, and the accuracy of this method is tested by experiments. The method uses Kinect device to capture bone tracking data of a number of people, and then determines the skeletal information of the user closest to Kinect in a certain range of areas in front of Kinect field of vision. The operator is determined by the coordinate information of the user's head and the sound source location function of the Kinect linear microphone array. The user's voice commands are captured and recognized in real time. The interaction between user and virtual assembly system is accomplished by simulating keyboard events.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN912.34

【参考文献】