【作者】 陈颖；

【导师】 吴长奇；

【作者基本信息】 燕山大学 ， 电路与系统， 2012， 硕士

【摘要】 基于麦克风阵列的声源定位技术是阵列信号处理领域的热点研究问题之一，有着广阔的应用前景。关于此项技术取得的科研成果多数是关于声源定位算法的研究，而对于声源定位算法的嵌入式系统开发则相对较少。为此，开展基于麦克风阵列的声源定位算法研究及实现具有重要的科研价值。本文侧重于对声源定位算法的实现，开发了一套基于TMS320F2812的声源定位系统。首先，研究了语音活动端点的检测，以避免无效的声源定位。给出了麦克风阵列近场和远场信号模型的划分依据，详细介绍了这两种信号模型的联系与区别。详细分析了麦克风阵列设计的要素，给出了阵列设计的一般原则。其次，介绍了基于麦克风阵列的时延估计声源定位方法，重点讨论了广义互相关时延估计法，并进一步改进了GCC-PHAT-ργ时延估计方法。此外还引入了一种相关峰细化精确时延估计方法，将这种算法与本文改进的算法进行了运算量的比较。之后，给出了基于时延参数的四阵元直线阵列的几何定位方法。再次，构建了以TMS320F2812为核心处理器件的实时声源定位系统。详细介绍了系统的硬件结构和软件程序设计。为了直观清楚的观察声源定位结果，开发了上位机软件显示界面。最后，对本文设计的声源定位系统进行实验测试，结果表明本文改进的算法具有更好的定位性能，该系统的定位成功率较高，且定位误差能够控制在一个较小的范围内，并对运动缓慢的移动声源具有一定的跟踪能力。更多还原

【Abstract】 Sound source localization based on microphone array is a new hotspot in the arraysignal processing, which has a wide range of application prospect. The majority ofachievements in scientific research about this technology are on the research of soundsource localization algorithm, but the development of embedded systems about thetechnology is relatively less. Therefore, it is of a great scientific value to research soundsource localization algorithms and realize it by microphone array. This paper focuses onthe realization of sound source localization algorithm, and develops a set of sound sourcelocalization system based on TMS320F2812.Firstly, this paper studies the detection of voice activity endpoint, which avoidsinvalid sound source localization. The article gives the divisiory basis of the microphonearray near filed and far filed signal model, and analyzes the relations and difference of thetwo kinds of signal model in detail. The design elements about the microphone array areanalyzed in detail, and the general principles of the design are given.Secondly, time delay estimation of sound source localization algorithm based onmicrophone array is introduced, and focuses on discussing generalized cross-correlationtime delay estimation algorithm, and further improves the method of GCC-PHAT-ργ.Besides, a kind of algorithm study of accurate time delay estimation with fineinterpolation of correlation peak is introduced, and compares the algorithm with theimproved algorithm on the computation. Then, the geometric positioning method of thefour elements linear array based on the parameter of time delay is given.Thirdly, a real-time sound source localization system is built, and TMS320F2812 isthe core processing component. This paper introduces the hardware structure of the systemand the software programming in detail. In order to observe the results of sound sourcelocalization directly and clearly, a software interface is developed.Finally, a great amount of experiments have been done to test the sound sourcelocalization system, and the results shows that the improved algorithm has betterperformance, and the system has a higher successful rate in positioning, and the error of localization can be controlled within a smaller scope, besides the system has certaintracking ability of sound source moving slowly.更多还原