基于中文语音交互的虚拟装配技术研究  

Research on Virtual Assembly Technology Based on Chinese Voice Interaction

作　　者：刘晓晖[1] 万丽莉[1,2] 季红[3] 苗振江[1] 

机构地区：[1]北京交通大学计算机与信息技术学院信息科学研究所,北京100044 [2]北京航空航天大学虚拟现实技术与系统国家重点实验室,北京100191 [3]北京市计算中心,北京100094

出　　处：《系统仿真学报》2014年第9期2056-2061,共6页Journal of System Simulation

基　　金：中央高校基本科研业务费专项资金(2014JBM027)

摘　　要：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。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,提出了一种基于中文语音指令交互辅助控制虚拟装配的方法,并设计了一套自行车虚拟装配系统,通过实验验证了方法的可行性。为了解决多人环境下的操作者确认问题,综合利用Kinect的骨骼跟踪和语音识别功能确定操作者,并识别其语音指令;根据零部件的物理连接关系和装配约束关系,定义了零件的层次化优先级模型,并在此基础上基于推荐规划装配序列,提高了易用性,降低了识别的错误率。At present, the force-feedback or somatosensory devices are usually utilized as interactive tools in virtual assembly systems, and some intuitive gestures, such as grab and drag, are recognized to complete the assembly of product. However, when some subtle parts need to be assembled, the results are not very desirable. In order to solve this problem, using Kinect as the input device, a virtual assembly method interacted by Chinese voice commands was proposed, and a bicycle virtual assembly system was applied. Experimental results demonstrate the ability of the proposed method. In order to confirm an operator in a multiplayer environment, skeleton tracking and speech recognition were both used to find the operator, and his voice commands were recognized. Due to physical connections and constraint relations, a hierarchical structure of part priority was also defined, and on this basis, the assembly sequence was planned and recommended to users, which could improve the usability and reduce the recognition error rate.

关 键 词：虚拟装配 KINECT 语音识别 装配序列规划 骨骼跟踪 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]