宽带波束形成器的抽头稀疏化设计研究
发布时间：2018-10-05 14:06
【摘要】：宽带波束形成是麦克风阵列信号处理的关键技术之一,在音频信号采集和声源定位等领域具有广泛的应用。基于滤波求和结构的空间响应变化约束最小二乘法是近年来新发展的频率不变宽带波束形成器(Frequency invariant beamformer)设计方法,对高性能宽带音频信号采集具有重要意义。而近年来所提出的基于Farrow结构的指向可调宽带波束形成器可以有效解决声源运动情况下的波束形成问题。本文以宽带波束形成器的抽头稀疏化设计为主线,分别对以上两类宽带波束形成器设计进行了比较深入的研究。主要研究工作包括:1.研究了基于滤波求和结构的空间响应变化约束最小二乘频率不变宽带波束形成器的抽头稀疏化设计。研究结果表明,该方法的性能受滤波器抽头长度的影响。通过增大滤波器的抽头长度,可以改善波束形成器的性能,但是其代价是波束形成器的实现复杂度亦增大。针对这一问题,本文提出了稀疏化设计方法,有效降低了频率不变宽带波束形成器的冗余权值。仿真结果表明,与相同抽头数量的非稀疏设计方法相比,基于稀疏化设计方法具有更好的性能。2.研究了基于Farrow结构的指向可调鲁棒宽带波束形成器的稀疏化设计。(1)针对麦克风失配误差有界的条件下,研究了基于最差性能优化的设计方法。该方法虽然已成功应用于基于滤波求和结构的波束形成器鲁棒设计,但研究结果表明,由于存在过约束问题,该方法并不适用于基于Farrow结构的指向可调宽带波束形成器设计。(2)针对麦克风失配误差的均值和方差的分布范围已知的条件下,研究了基于最差平均性能优化和通带响应方差约束的鲁棒设计方法。研究结果表明,当存在麦克风失配误差时,该方法具有良好的性能。此外,还分析了基于Farrow结构的指向可调宽带波束形成器的通带响应方差的一些特性。(3)研究了基于Farrow结构的指向可调宽带鲁棒波束形成器的稀疏化设计。与经典的滤波求和结构波束形成器相比,基于Farrow结构的波束形成器指向可调的实现是以增加滤波器的个数为代价的。提出了一种基于Farrow结构的指向可调宽带鲁棒波束形成器的抽头稀疏化方法,有效剔除了波束形成器的冗余抽头权值,降低了实现复杂度。仿真结果表明,与相同抽头个数的非稀疏波束形成器相比,稀疏化设计有效改善了波束形成器的性能。
[Abstract]:Broadband beamforming is one of the key technologies of microphone array signal processing. It has been widely used in audio signal acquisition and sound source location. Spatial response constrained least square method based on filter summation structure is a new design method of frequency invariant wideband beamformer (Frequency invariant beamformer) which is of great significance for high performance wideband audio signal acquisition. In recent years, the Farrow structure based wideband beamformer can effectively solve the problem of beamforming when the sound source is moving. Based on the thinning design of wideband beamformer, the design of these two kinds of wideband beamformer is studied in detail in this paper. The main research work includes: 1. The thinning design of wideband beamformer with variable spatial response constraints and frequency invariant least-squares based on filter summation structure is studied. The results show that the performance of the method is affected by the filter tap length. By increasing the tap length of the filter, the performance of the beamformer can be improved, but the cost is that the implementation complexity of the beamformer is also increased. To solve this problem, a sparse design method is proposed to effectively reduce the redundant weights of frequency invariant broadband beamformer. The simulation results show that the sparse design method has better performance than the non-sparse design method with the same number of taps. The thinning design of directionally adjustable robust wideband beamformer based on Farrow structure is studied. (1) under the condition that microphone mismatch error is bounded, the design method based on worst-case performance optimization is studied. Although the proposed method has been successfully applied to the robust design of beamformer based on filter summation structure, the results of the study show that there are overconstrained problems. This method is not suitable for the design of directionally adjustable wideband beamformer based on Farrow structure. (2) when the distribution range of mean and variance of microphone mismatch error is known, A robust design method based on the worst average performance optimization and passband response variance constraints is studied. The results show that the method has good performance when microphone mismatch error exists. In addition, some characteristics of passband response variance of directional tunable wideband beamformer based on Farrow structure are analyzed. (3) sparseness design of directional tunable wideband robust beamformer based on Farrow structure is studied. Compared with the classical filter summation beamformer, the implementation of Farrow based beamformer with adjustable direction is at the cost of increasing the number of filters. In this paper, a tap thinning method based on Farrow structure for directional tunable wideband robust beamformer is proposed, which effectively eliminates the redundant tap weights of beamformer and reduces the complexity of implementation. The simulation results show that the sparse design improves the performance of the beamformer effectively compared with the non-sparse beamformer with the same number of taps.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.7
，
本文编号：2253687
[Abstract]:Broadband beamforming is one of the key technologies of microphone array signal processing. It has been widely used in audio signal acquisition and sound source location. Spatial response constrained least square method based on filter summation structure is a new design method of frequency invariant wideband beamformer (Frequency invariant beamformer) which is of great significance for high performance wideband audio signal acquisition. In recent years, the Farrow structure based wideband beamformer can effectively solve the problem of beamforming when the sound source is moving. Based on the thinning design of wideband beamformer, the design of these two kinds of wideband beamformer is studied in detail in this paper. The main research work includes: 1. The thinning design of wideband beamformer with variable spatial response constraints and frequency invariant least-squares based on filter summation structure is studied. The results show that the performance of the method is affected by the filter tap length. By increasing the tap length of the filter, the performance of the beamformer can be improved, but the cost is that the implementation complexity of the beamformer is also increased. To solve this problem, a sparse design method is proposed to effectively reduce the redundant weights of frequency invariant broadband beamformer. The simulation results show that the sparse design method has better performance than the non-sparse design method with the same number of taps. The thinning design of directionally adjustable robust wideband beamformer based on Farrow structure is studied. (1) under the condition that microphone mismatch error is bounded, the design method based on worst-case performance optimization is studied. Although the proposed method has been successfully applied to the robust design of beamformer based on filter summation structure, the results of the study show that there are overconstrained problems. This method is not suitable for the design of directionally adjustable wideband beamformer based on Farrow structure. (2) when the distribution range of mean and variance of microphone mismatch error is known, A robust design method based on the worst average performance optimization and passband response variance constraints is studied. The results show that the method has good performance when microphone mismatch error exists. In addition, some characteristics of passband response variance of directional tunable wideband beamformer based on Farrow structure are analyzed. (3) sparseness design of directional tunable wideband robust beamformer based on Farrow structure is studied. Compared with the classical filter summation beamformer, the implementation of Farrow based beamformer with adjustable direction is at the cost of increasing the number of filters. In this paper, a tap thinning method based on Farrow structure for directional tunable wideband robust beamformer is proposed, which effectively eliminates the redundant tap weights of beamformer and reduces the complexity of implementation. The simulation results show that the sparse design improves the performance of the beamformer effectively compared with the non-sparse beamformer with the same number of taps.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.7
，
本文编号：2253687
本文链接：https://www.wllwen.com/kejilunwen/wltx/2253687.html
