摘要语音通信作为人类最重要的信息交流方式之一，其通信质量极易受到外界环境和 传输媒介的影响，从而不仅降低信息获取质量而且恶化了语音处理系统性能。波束形 成的实质是对阵列采集到的声场信号进行线性加权求和处理，使信号聚焦于期望方向 输出，达到抑制其他方向干扰和噪声的目的，因此波束形成广泛地应用于在声源识别、 语音增强、声源定位等领域。 

本研究结合球阵列的对称结构和球谐分解特性，在传统波束形成技术的理论基础 上，提出基于球阵列的 MVDR(minimum variance distortionless response，最小方差无 失真响应)波束形成技术。同时，考虑到复值信号模型的计算复杂性，本研究在此技 术上提出一种实值变换思想，即在不损失阵列接收信号信息的前提下，将复值信号模 型通过一定方式转化为实值情况。本研究以 32 元均匀刚性球阵列为例进行仿真，实 验结果表明，较之以前的复值算法，球阵列的实值 MVDR 波束形成技术既可在阵列 输出波束图、估计声源方向上具有良好性能，又可很大程度地降低计算量。 

因此，本研究提出的该算法对信号的处理表现出明显成效，使其在声源定位、语 音增强等语音信号处理领域中具有重要意义。 78332

毕业论文关键词：球阵列；实值 MVDR 波束形成；实值变换 

AbstractVoice communication acts as one of the most important ways of information communication for humans and its quality can be easily influenced by the external environment and the transmission medium。 It not only reduces the quality of voice communication and also deteriorates the overall performance of the voice communication system。 The essence of beamforming is to make a linear weighted sum treatment of the acoustic signal captured by the array so that to focus on the desired direction of the signal output and achieve the goal of suppression of unwanted signals in other directions。 Therefore, beamforming is widely applied in sound source recognition, speech enhancement, and sound source localization。

With the properties of symmetric structure and spherical harmonic decomposition of spherical arrays and based on the theory of traditional beamforming, this paper proposes a MVDR (minimum variance distortionless response) beamforming technology of spherical arrays。 Meanwhile, considering computational complexity of complex-valued signal model, an idea of real-valued transform is presented。 That is to say, the complex-valued signal model is transformed into a real-valued one by some processes without  sacrificing receiving signal information of arrays。 This paper takes 32 uniform rigid spherical arrays as an example to carry out simulation。 The experimental results show that compared with the complex-valued method, the real-valued MVDR beamforming algorithm of spherical arrays not only can have a fine performance on the beampattern of signals and the estimation of the incident direction of the sound source, but also can reduce the amount of computation to a great extent。

Therefore, the proposed algorithm for signal processing shows significant results and it is of great significance in sound localization, speech enhancement and other speech signal processing field。

Key words: spherical arrays; real-valued MVDR beamforming; real-valued transform

目 录

第一章  绪论 - 1

1。1 研究背景及意义 - 1 

1。2 波束形成技术国内外研究现状 - 2 

1。3 本文主要研究内容章节安排 - 3 

第二章  阵列信号处理基本理论 - 5

2。1 阵列信号处理概念 - 5 

2。2 阵列拓扑结构 - 5 

2。2。1 常见阵列结构 - 5