摘要随着科技的发展，多通道声源分离技术作为研究的重要性越来越凸显出来，即如何把不同的目标声音在混合的声音中去有效地分离出来，这项研究已经在通信、源目标检测、声学信号分析等方面都发挥了作用。25847
本文针对从混合观测声音数据中恢复出不可直接观测的各个声源信号分离问题，研究了基于麦克风阵列的多通道声源分离技术原理。并使用麦克风阵列开展实测室内多声源混合信号的采集、分析与处理方法。对基于相位差（IPD）和幅度差（ILD）特征的声源分离技术进行分析，并比较二者在高低频段以及不同阵列结构下的表现。研究结合相位差和幅度差互相弥补的特点，根据人类双耳听觉定位机理研究了模拟该机理的相位差-幅度差融合声源分离技术，最后Matlab软件对采集来的语音数据进行处理和分析，并对其图表结果进行详细描述和合理物理解读。实验结果以及仿真证明了相位差 –幅度差差融合的源分离方法的有效性。
关键词 声源分离 线性麦克风阵列 相位差 幅度差 聚类算法
毕业设计说明书（毕业论文）外文摘要
Title Phase and Level Fusion For Robust Multichannel Source Separation
Abstract
In recent years, the multichannel source separation has been an important topic. That is how to effectively separate the different target sound in the mixed voice, The research has important theoretical and practical value on voice communication, acoustic target detection.
In this paper, the problem of separating the acoustic source signals from the mixed observation sound data is recovered. multichannel source separation based on microphone array is studied，And carry out experiment analysis and processing of mixed signal acquisition, multi sound source experiment with the microphone array. It is analyzed that the common features Inter-channel phase (IPD) and level (ILD) by multichannel source separation. And compare the performance of two in different frequency and different structure of microphone array. Study the characteristics of the combination of complementary Inter-channel phase (IPD) and level (ILD). According to the human binaural auditory localization, Simulation mechanism of dual microphone phase of the IPD-ILD fusion sound source separation technology. Finally, the MATLAB software is processed and analyzed to the speech data collected. And give a detailed description and interpretation of the reasonable physical chart. Experimental results and simulations demonstrate the effectiveness of IPD-ILD fusion sound source separation technology.
Keywords source separation、microphone array、 IPD、 ILD、 RANASC
目 次
1 绪论 1
1.1 课题研究意义与目的 1
1.2 国内外的发展现状及趋势 2
1.3 本论文完成的工作 2
1.4 本文的主要内容与结构 2
2 麦克风阵列分析原理 4
2.1 语音信号模型 4
2.2 麦克风阵列信号模型 5
2.3 麦克风阵列语音信号处理 7
3 多通道声源分离技术原理 10
3.1 人耳听觉定位基本原理 10
3.2 音频频道间的特征 11
3.3 聚类算法 13
3.4 源分离 15
4 数据采集及预处理 17
4.1 声信号采集 17 基于幅度差及相位差融合的多通道声源分离技术研究:http://www.751com.cn/tongxin/lunwen_19815.html