本文为美国马奎特大学（作者：HeatherElaine Ewalt）的硕士论文，共125页。

本文介绍了一种采用麦克风阵列波束形成和语音增强算法的系统设计与实现。该系统的目标是提高主语音信号的质量，波束形成器的工作方式是通过利用阵列信号信息而不是物理移动阵列，将一组麦克风转向所需要的观察方向。通过最小化非视线方向上的干扰源和噪声能量，同时增强视线方向上信号的能量来实现这一目的。本文研究了两种波束形成方法：延迟求和DS波束形成器和最小方差无失真响应MVDR波束形成器。首先将输入信号分解为多个频带，以便采用窄带波束形成技术。两种波束形成方法分别采用多源维纳滤波和多源谱减增强算法，这些算法利用从初始波束形成算法获得的每个信号源的信号估计作为输入，这些多源增强算法可以通过迭代技术实现，以改进信号估计效果，同时提高主源的信噪比。

本文提出的实验装置由两个和三个使用线性麦克风输入系统的语音源组成，该算法既适用于模拟实验装置，也适用于语音处理的室内数据采集。为了衡量增强后语音信号质量的提高，对原始信号、波束形成信号和增强信号进行了整体信噪比和分段信噪比的测量比较。除了这些质量改进指标之外，还进行了听众意见的主观测试。

This thesis describes the design andimplementation of a speech enhancement system that uses microphone arraybeamforming and speech enhancement algorithms applied to a speech signal in amultiple source environment. The goal of the system is to improve the qualityof the primary speech signal. Beamformers work by means of steering an array ofmicrophones towards a desired look direction through utilizing signalinformation rather than physically moving the array. They accomplish this throughminimizing the energy of interference sources and noise in non-look directionswhile increasing the energy of the signal in the look direction. In thisresearch, two beamforming methods are examined: the delay and sum (DS)beamformer and the minimum variance distortionless response (MVDR) beamformer.The input signals are first split into frequency bands so that narrowbandbeamforming techniques can be used. Multiple source Wiener filtering andmultiple source spectral subtraction enhancement algorithms are incorporatedinto the two methods of beamforming. The algorithms utilize signal estimates ofeach source obtained from the initial beamforming algorithms as inputs. Thesemultiple source enhancement algorithms result in iterative techniques to improvethose estimates while improving the signal to noise ratio of the primarysource.

The experimental setup presented hereconsists of both two and three speech sources using a linear microphone inputsystem. The algorithms are performed on both simulated experimental setups andon data obtained from a data acquisition system in an acoustically treatedsound room. To measure the improvement in quality of the enhanced signal,overall SNR and segmental SNR improvement is determined for the original,beamformed, and enhanced signal. In addition to these quality improvementmetrics, listener opinion testing is performed.

1 引言

1.1 论文声明

1.2 论文概述

2 项目背景

2.1 麦克风阵列基础

2.2 波束形成器基础

2.3 波束形成器的具体实现

2.4 语音增强基础

2.5 语音增强的测量基础

3 迭代的多源增强方法

3.1 多源谱减增强

3.2 多源维纳滤波增强

3.3 耦合函数

4 实验设置

4.1 实验设备

4.2 多个说话者的输入信号

4.3 算法处理细节

5 数据采集系统

5.1 多个说话者的输出系统

5.2 多输入系统

5.3 音响设置

6 实验结果

7 讨论

8 结论

附录A 模拟数据的实验结果

附录B MOS测试

更多精彩文章请关注公众号：