中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
基于传声器阵列的说话人定位研究

文献类型:学位论文

作者张倩
学位类别博士
答辩日期2009-05-14
授予单位中国科学院声学研究所
授予地点声学研究所
关键词传声器阵列 声源定位 时延估计 自适应特征值分解 α-β 滤波器
其他题名Research of Speaker Localization Based on Microphone Arrays
学位专业信号与信息处理
中文摘要基于传声器阵列的定位技术是一项有广阔应用前景的研究,尤其在与人们日常生活紧密相关的一些领域更是取得了广泛的关注。基于传声器阵列的说话人定位技术在视频会议、语音检测等领域有重要的应用价值。但由于噪声和混响的存在,使这项研究遇到一定的困难,使得很多方法都受到了使用上的限制。 本文系统地研究了基于传声器阵列的说话人定位方法。首先对传声器阵列及其研究现状进行了总体概述,讨论了基于传声器阵列的声源定位所面临的问题,分析了阵列信号处理的特殊性和混响,噪声的产生原因及影响。总结归纳并比较了各种基于传声器阵列声源定位方法的优缺点。展开分析了基于时延估计的声源定位方法。在详细讨论几种时延估计方法后,针对其中的自适应特征值分解法进行了改进,提高其收敛能力。然后介绍了几种常用的空间定位法,包括角度距离定位法、球形插值法及线性插值法等。 最后利用实验室现有的条件搭建了实际的说话人定位系统,采用了改进的自适应特征值分解法,并采取了α-β滤波器对数据进行简单的后置处理。从结果看该系统能够较好的抑制混响和噪声的影响,并能较快收敛到稳定时延值。文中给出了该系统的实现框图及数据处理的结果,并对误差进行了详细的分析,验证了系统的可行性。
英文摘要Localization of sound source is an important study which will be widely applied in our future life. Especially for its use to daily life, people pay more and more attention to this technique. Microphone array can be employed for speaker localization in videoconference and voice detection. However, background noise and room reverberations in enclosure greatly degrade the effectiveness of acoustic source localization. Some methods are also inhibited because of this phenomenon. The methods for speaker localization using microphone array are studied in this paper. First, the background of research of microphone arrays is described and problems in this area are discussed. The particularities of array signal processing, and the influence of noise and reverberation are analyzed, too. Some main methods of sound source localization are compared and sound source localization methods based on the time delay estimation(TDE) are emphasized. After analyzing three TDE methods a approach to improve the convergence of adaptive eigenvalue decomposition(AED) algorithm is proposed in this paper. Then several locating methods including angle-distance method, spherical interpolation, linear interpolation are introduced. Finally, a speaker localization experiment system is established based on available hardwares and softwares. Because of proper method of TDE and α-β filter to smoothing the results, this system is tested to be robust to noise and reverberation and the algorithm performs better convergence. This paper gives the implementation and result of data processing, and detailed error analysis. The feasibility of the system is testified.
语种中文
公开日期2011-05-07
页码66
源URL[http://159.226.59.140/handle/311008/574]  
专题声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文
推荐引用方式
GB/T 7714
张倩. 基于传声器阵列的说话人定位研究[D]. 声学研究所. 中国科学院声学研究所. 2009.

入库方式: OAI收割

来源:声学研究所

浏览0
下载0
收藏0
其他版本

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。