单位代码： 10293 密 级： 公开
fromdust专 业 学 位 硕 士 论 文
dragonboatfestival
论文题目：麦克风阵列波束成形算法研究与实现
Rearch and Implementation of Microphone Array Beamforming Algorithm
Thesis Submitted to Nanjing University of Posts and Telecommunications for the Degree of
Master of Engineering
女厕所英文
By
Chen YingRui
Supervisor: Prof. Wu Meng
May 2020brooklyn
南京邮电大学学位论文原创性声明
本人声明所呈交的学位论文是我个人在导师指导下进行的研究工作及取得的研究成果。尽我所知，除
了文中特别加以标注和致谢的地方外，论文中不包含其他人已经发表或撰写过的研究成果，也不包含为获得南京邮电大学或其它教育机构的学位或证书而使用过的材料。与我一同工作的对本研究所做的任何贡献均已在论文中作了明确的说明并表示了谢意。 本人学位论文及涉及相关资料若有不实，愿意承担一切相关的法律责任。
南京邮电大学学位论文使用授权声明
本人承诺所呈交的学位论文不涉及任何国家秘密，本人及导师为本论文的涉密责任并列第一责任人。
本人授权南京邮电大学可以保留并向国家有关部门或机构送交论文的复印件和电子文档；允许论文被查阅和借阅；可以将学位论文的全部或部分内容编入有关数据库进行检索；可以采用影印、缩印或扫描等复制手段保存、汇编本学位论文。本文电子文档的内容和纸质论文的内容相一致。论文的公布（包括刊登）授权南京邮电大学研究生院办理。 非国家秘密类涉密学位论文在解密后适用本授权书。
so as to
研究生学号：_1217012135_ 研究生签名： 日期：美剧排行榜前十名
_2020.5.13
研究生签名： 导师签名： _ 日期： 2020.5.13
摘要
随着语音技术的迅速发展，越来越多的用户有意获取更为清晰的语音系统。在视频、电话会议系统、大型会议室等一些应用场合，需要对感兴趣的声源实现定位和语音增强。传统的单个麦克风在对采集到的语音进行增强算法处理时，容易混入新的音乐噪声且造成信号的失真，不能满足人们对声音质量的要求。麦克风阵列能够自动检测，定位声源的位置，并对噪声信号进行空间滤波，取得更加明显的干扰噪声增强语音的效果。the wizard of oz
本论文主要研究了基于麦克风阵列波束形成的声源定位算法、语音增强算法以及回声与混响消除算法。论文研究了基于高分辨率谱估计的声源定位技术、基于最大输出功率的可控波束形成定位技术等几种典型的声源定位算法，基于到达时间差（TDOA）定位技术提出了一种改进的定位技术，改进后的声源定位算法能够更加准确地识别声源的位置，提高时延估计的准确性。研究了常规波束形成算法、自适应波束形成算法和后置滤波器波束形成算法，基于后置滤波器法设计出一种基于自适应频谱
降噪（ASNR）和多源选择（MSS）的后置滤波器波束形成算法，对算法的性能进行了比对分析。其中常规波束形成法在一定程度上能够实现语音增强，它去除噪声的效果有限且只能消除相干噪声；GSC算法处理的过程中，滤除期望信号估计噪声成分的能力不是很好，影响算法整体的消噪性能；本文提出的基于自适应频谱降噪（ASNR）和多源选择（MSS）的后置滤波器波束形成算法能减少来自中频段和高频段的噪声，也能利用自适应ASNR滤波器减少低频噪声的影响，提高了消噪的性能以及输出音频信号的质量。研究了麦克风阵列混响消除算法与自适应回声消除算法，自适应回声消除算法通过调整自适应滤波器的权值，模拟出一个无限逼近真实回声路径的近似回声路径，得出估计的回声信号，对算法的性能进行了仿真，仿真结果表明使用自适应回声消除算法能够消除语音信号中回声成分，处理效果良好。
重庆特训营关键词: 麦克风阵列，波束形成，声源定位，语音增强，C6747，OMAP-L137
Abstract
With the rapid development of voice technology, more and more urs intend to obtain a clearer voice system. In some applications such as video, teleconferencing systems, and large conference rooms, it is necessary to implement localization and voice enhancement for sound sources of interest. The traditional single microphone performs enhanced algorithm processing on the collected
linq
voice, which easily enters new music noi and caus signal distortion, cannot meet people's requirements for sound quality. The microphone array can automatically detect, locate the sound source, and spatially filter the noi signal to obtain a more obvious effect of enhancing speech by inhibiting noi.
舒婷我亲爱的祖国This thesis mainly studies the sound source localization algorithm, speech enhancement algorithm and echo and reverberation cancellation algorithm bad on microphone array beamforming. The paper studies veral typical sound source localization algorithms such as high-resolution spectral estimation of sound source localization technology, controllable waveform formation localization technology bad on the maximum output power, and propos an improved localization technology bad on the time of arrival (TDOA). The improved sound source localization algorithm can more accurately identify the position of the sound source and improve the accuracy of the delay estimation. The Convention Beamforming algorithms, the Adaptive Beamforming algorithms and the post-filter beamforming algorithm are studied, and a post-filter beamforming algorithm bad on ASNR and MSS is designed. The paper compares and analyzes the performance of veral algorithms. Among them, the Convention Beamforming method can achieve speech enhancement to a certain extend. It has a limited effect of removing noi and can only remove coherent noi. In the
process of the GSC algorithm, the ability to filter out the noi component of the expected signal is not very good, affecting the overall algorithm. Post-filter beamforming algorithm bad on adaptive spectral noi reduction (ASNR) and multi-source lection (MSS) propod in this paper can reduce noi from mid-band and higher bands, and can also u adaptive ASNR filtering to reduce the influence of low-frequency noi, improve the performance of noi cancellation and the quality of the output audio signal. The paper studies the microphone array reverberation cancellation algorithm and adaptive echo cancellation algorithm. The adaptive echo cancellation algorithm simulates an approximate echo path that infinitely approximates the true echo path by adjusting the weight of the adaptive filter to obtain an estimated
