基于麦克风阵列的移动声源定位研究

发布时间：2018-12-08 19:04

【摘要】：近年来,利用麦克风阵列对信号源的角度和位置进行估计已成为一个热门的研究方向。在音视频会议摄像头自动循声定位、助听器声音去噪增强、智能机器人声音跟踪等方面已经有了大量的应用,但对移动声源的轨迹定位研究尚不多见。本文结合应用广泛的基于到达时延差(TDOA)和高分辨率谱估计(MUSIC)算法,对移动声源的轨迹定位进行研究,估计出声源随时间变化的角度轨迹。首先,研究了三种基本的定位方法,介绍了麦克风阵列结构及其信号模型,引入一种适用于运动声源定位的均匀直线阵列模型。其次,分析比较了几种不同的时延估计方法,并就其中的广义互相关算法针对不同加权函数做了仿真。对不同加权函数的抗噪声性能做了分析,重点讨论了PHAT-GCC算法。再次,对子空间技术中的高分辨率谱估计(MUSIC)算法的基本理论进行了研究,针对协方差矩阵特征值分解的复杂性,采用了一种免特征值算法,在保持谱估计精度的前提下,提高算法估计效率,并做了仿真验证。最后,针对移动声源轨迹定位的阵列信号模型,使用PHAT-GCC广义互相关和免特征值分解的MUSIC两种算法进行仿真,估计了声源在运动过程中俯仰角的变化轨迹。仿真分析了信噪比、采样率、和运动速度给定位系统带来的影响,并针对移动给定位系统带来的频率偏移问题进行了修正。
[Abstract]:In recent years, using microphone array to estimate the angle and position of signal source has become a hot research direction. There have been a lot of applications in audio and video conference camera, such as automatic tracking of audio and video conference camera, enhancement of sound denoising of hearing aid, sound tracking of intelligent robot and so on. However, the research on track location of moving sound source is still rare. In this paper, based on the widely used (TDOA) and high-resolution spectral estimation (MUSIC) algorithm, the track location of mobile sound source is studied, and the angle locus of sound source changing with time is estimated. Firstly, three basic localization methods are studied, the microphone array structure and its signal model are introduced, and a uniform linear array model for moving sound source location is introduced. Secondly, several different time delay estimation methods are analyzed and compared, and the generalized cross-correlation algorithms are simulated for different weighting functions. The anti-noise performance of different weighting functions is analyzed, and the PHAT-GCC algorithm is discussed emphatically. Thirdly, the basic theory of high-resolution spectral estimation (MUSIC) algorithm in subspace technology is studied. In view of the complexity of eigenvalue decomposition of covariance matrix, a no-eigenvalue algorithm is used to maintain the accuracy of spectral estimation. The efficiency of estimation is improved, and the simulation is done. Finally, aiming at the array signal model of moving sound source locus location, two kinds of algorithms, PHAT-GCC generalized cross-correlation and MUSIC with no eigenvalue decomposition, are used to simulate and estimate the track of pitch angle of sound source in the process of motion. The influence of signal to noise ratio, sampling rate and moving velocity on the positioning system is analyzed by simulation. The frequency offset problem brought by movement to the location system is corrected.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN642;TN911.7

【参考文献】