摘要随着声源处理越来越被重视，对声源的精确定位已成为众多学者的研究对象，并广泛应用于军事、科研、生活、工业等领域。本文主要研究的是在多源的情况下利用双麦克风阵列进行目标声源角度定位的方法，采用的是基于互功率谱相位时延差的测角方法，将时域中不易分离的声源信号转化为时频域中易于分离的信号。该方法先对声源信号取傅里叶变换，得到各声源信号的互功率谱，再利用互功率谱的相角信息得到时延值，由时延值大小将采样点分类。分类后将各类中采样点的互谱幅值聚类加权得到功率峰，根据功率峰便可分离信号源。最后利用一维到达时延差进行反正弦变换即可获得语音源的角度信息。

毕业论文关键词  麦克风阵列  多源定位   互谱加权  时延

毕业设计说明书（论文）外文摘要

Title   Multisource localization technology research based on the microphone array via cross-spectral clustering     

Abstract With the sound source processing more and more attention, the precise positioning of the sound source has become the object of study of many scholars, and is widely used in military, scientific research, life, industry and other fields. This paper studies the use of dual-microphone array for target sound source angle positioning method used is based on cross-power spectrum phase delay difference angle measurement method, easily separated in the time domain signal into time-frequency sound source domain easy to separate signals. This method is first to take the Fourier transform of the source signal, the sound source signals to obtain the cross power spectrum, and then cross-power spectrum using the phase angle information is delay value, the delay value of the sampling point size classification. Classification of sampling points will be all kinds of cross-spectral amplitude weighted clustering get power peak, the peak can be separated according to the power source. Finally, one-dimensional differential delay arrival arcsine transformation can be obtained for speech source angle information.

Keywords  Microphone array  Multisource localization   Cross spectrum weighted  Delay

1  引言 1

1.1  研究背景和意义 1

1.2  国内外研究的概况及发展趋势 2

1.3  本文的主要内容及构成 3

2  麦克风阵列基本测向原理 3

2.1  基于到达时间差的测向原理 3

2.2  基于空间谱估计（DOA）的定位原理 5

3  语音信号的互谱聚类算法 6

3.1  常用的时延估计法思想 6

3.1.1  广义互相关法 6

3.1.2  最小均方自适应滤波法 7

3.1.3  互谱加权时延估计思想 7

3.2  互谱聚类麦克风阵列的测向算法 8

4  实验仿真及结果分析 10

4.1  实测数据分析 12

4.2  仿真操作说明 20

4.3 误差来源分析 20

5 实验遇到的问题及解决方法 21

5.1  声音采集 21