移动机器人语音定向算法及其实现  

Speech Localization Algorithm forMobile Robots and Its Application

作　　者：靳晓强[1,2] 张汝波 刘冠群 

机构地区：[1]中国核科技信息与经济研究院,北京100037 [2]哈尔滨工程大学计算机科学与技术学院,黑龙江哈尔滨150001

出　　处：《计算机仿真》2012年第11期223-226,255,共5页Computer Simulation

基　　金：国家高技术研究发展计划(863计划)(2009AA04Z215)

摘　　要：研究在光线比较差或者有障碍物的复杂环境下对感兴趣的声源进行定位,由于存在外部噪声,引起对目标声源进行跟踪误差大。传统的语音定向方法无法有效减小声源定位的误差,从而定向的精度比较低,且多集中在理论研究。为提高移动机器人语音定向的精度,利用麦克风阵列拾取声源信号,采用PC/104的SEM/ADT800多通道数据采集卡对语音信号进行采集和处理。运用改进的广义互相关时延估计算法计算麦克风阵元间的时间差值,采用四元十字阵形的几何定位方法计算声源信号的目标位置和方向,并对系统的定向效果进行实验,得出实验结果,进行误差分析。经实验证明,相对于传统的语音定向方法,麦克风阵列的移动机器人语音定向系统实际可用,具有较好的性能,能够把系统定向误差控制在要求的范围内。Study the problems in complex environment, such as poor light or obstructions, locating the interesting sound source, and even tracking the target sound source. The traditional methods can not effectively reduce the sound source localization error, the accuracy is relatively low, and most studies focus on the theory. To improve the accura- cy of speech localization on a mobile robot, the microphone array was used to pick up sound source signals, SEM / ADT800 multi-channel data acquisition card based on PC/104 was applied for the speech signal acquisition and pro- cessing. This paper employed the generalized cross-correlation time delay estimation algorithm to calculate the time difference between the microphone array elements. Four-element cross Lineup geometric positioning was used to cal- culate the sound source signal target location and orientation. Then, it tested the speech localization system, gave the experiment results, and analyzed the errors. The experiment proves that the speech localization system based on mi- crophone array on a mobile robot is actually available, with good performance compared with the prediction method, and is able to control the system error in the required range.

关 键 词：语音定向 数据采集 时延估计 互相关 麦克风阵列 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]