基于宽间距麦克风阵列的多声源测向方法研究

发布时间：2018-01-15 15:03

本文关键词：基于宽间距麦克风阵列的多声源测向方法研究　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：本课题依托国家自然科学基金(61171167,61401203),开展了对宽间距麦克风阵列多声源测向方法研究。基于麦克风阵列的多声源测向技术通过处理由阵列接收的混合声频信号来确定各个声源的方位,在军事和民用的多个领域有着广泛应用。例如在视频会议中,利用该技术精确估计发言人所在方位可以对其语音进行增强并准确控制摄像头。通过接收和处理枪炮声信号,还能实现敌方狙击手探测定位以及战场态势检测等功能。此外,随着对声源信号分析带宽要求和空间分辨率要求的不断提高,采用宽间距麦克风阵列获取更宽频带信号信息是有效控制实际声测系统复杂度和成本的合理措施。因此对宽间距麦克风阵列多声源测向方法的研究将具有十分重要的意义。本文首先归纳总结了传统的声源测向方法。由于基于时延估计的波达方向估计技术具有计算量小、精确性高等优势,本文主要讨论该领域最基础问题之一的双元麦克风阵列时延估计技术。针对远场多声源模型,在分析比较传统广义互相关(GCC)法和退化分离估计技术(DUET)的特点及局限性基础上重点对角度谱类方法开展研究,分别说明了基于最小方差无失真响应(MVDR)和扩散噪声模型(DNM)的两种聚合角度谱,并对各自特点进行了分析比较。该类方法能够在混响环境下实现多声源测向,但是当声源信号存在波长之半小于麦克风间距的高频成分时,将会由于空间混叠出现相位卷绕模糊问题,限制了其在宽间距阵列中的应用。针对上述问题,本文研究了一种新的聚合角度谱方法,采用声源能量积累策略和恒虚警率检测技术,仅对频域积累输出中的高值分量进行帧间时域聚合处理,显著降低了聚合角度谱中的空间混叠伪峰高度和数量。详细阐述了该方法的原理及实现流程,并开展了大量仿真和外场实验验证。实验结果表明,本文所述方法在应用于宽间距阵列时能够显著抑制空间混叠的影响,有利于提高多声源检测和测向性能。
[Abstract]:This subject relies on the National Natural Science Foundation of China 61171167 (61401203). The multi-source direction-finding technology based on microphone array determines the azimuth of each sound source by processing the mixed audio signals received by the array. It has a wide range of applications in military and civilian fields, such as videoconferencing. Using this technique, the speaker's voice can be enhanced and the camera can be accurately controlled by accurately estimating the location of the speaker, and the sound signal of the gun can be received and processed. In addition, with the increasing requirement of bandwidth and spatial resolution for the analysis of sound source signal, the detection and location of enemy sniper and battlefield situation can be realized. It is a reasonable measure to control the complexity and cost of practical sound measurement system by using wide spacing microphone array to obtain wider band signal information. Therefore, the research on direction finding method of multi sound source with wide spacing microphone array will be very important. This paper first summarizes the traditional acoustic source direction finding methods. Due to the time delay estimation based on the direction of arrival estimation technology has a small amount of calculation. In this paper, one of the most basic problems in this field, the dual-element microphone array time delay estimation technology, aimed at the far-field multi-source model. Based on the analysis and comparison of the characteristics and limitations of the traditional generalized cross-correlation GCCs and the degenerate separation estimation techniques, this paper focuses on the study of angle spectrum methods. Two convergent angle spectra based on minimum variance distortion free response (MVDR) and diffusion noise model (DNM) are presented respectively. This method can realize multi-source direction finding in reverberation environment, but when the sound source signal has a high frequency component of half wavelength less than microphone spacing. Due to the phase winding ambiguity problem in spatial aliasing, its application in wide spaced array is limited. In view of the above problems, a new method of polymerizing angle spectrum is studied in this paper. Based on the acoustical source energy accumulation strategy and CFAR detection technique, only the high value components in the frequency domain accumulation output are processed by time domain aggregation. The height and the number of pseudo peaks of spatial aliasing in the polymerization angle spectrum are significantly reduced. The principle and implementation flow of this method are described in detail, and a large number of simulations and field experiments are carried out. The experimental results show that. The method proposed in this paper can significantly suppress the influence of spatial aliasing when it is applied to wide spacing arrays, and is helpful to improve the detection and direction finding performance of multi-sound sources.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN912.3

【参考文献】