麦克风阵列近场波束形成典型方法的比较研究

发布时间：2018-05-02 13:07

本文选题：麦克风阵列 + 近场　；参考：《内蒙古大学》2017年硕士论文

【摘要】：波束形成是麦克风阵列信号处理的核心技术之一,沿用并发展了阵列信号处理的方法。传统的波束形成方法多数基于远场窄带信号处理,适用于目标信号处于阵列远场范围内的情况。当麦克风阵列系统应用于近场环境时,如小型会议室、车内免提电话等,若用远场信号处理模型,则会导致信号处理的效果明显下降。此外,在麦克风阵列的近场应用环境中,目标信号来源于语音等宽带信号。因此,对麦克风阵列近场窄带和宽带波束形成方法的研究更符合麦克风阵列系统的实际应用场景。本文主要对基于近场球面波模型的麦克风阵列波束形成方法进行了研究。首先,研究了近场信号处理模型,根据近场信号传播模型,得到近场阵列响应矩阵,由此得到均匀线阵近场导向矢量。和远场阵列响应矩阵相比,近场阵列响应矩阵随方位和距离而变化,而远场阵列响应矩阵只受到方位变化的影响。其次,研究了近场窄带波束形成方法,主要包括常规波束形成、最小方差无畸变响应(Minimum Variance Distortionless Response,MVDR)波束形成、最小均方误差(Least Mean Square,LMS)算法、递推最小二乘(Recursive Least Mean Square,RLS)算法和采样矩阵求逆(Sample Matrix Inversion,SMI)算法。将这几种窄带波束形成方法基于近场均匀线阵进行仿真实验,并对它们在近场环境下的分辨能力、抗干扰能力、运算速度、抑制干扰信号能力等性能进行了比较。最后,研究了近场宽带波束形成方法。针对宽带信号处理的特殊性,介绍了三种波束形成方法,包括基于时域信号处理的常规时延波束形成、基于频域分解的非相干信号子空间处理方法(Incoherent Signal-subspace Method,ISM)和相干信号子空间处理方法(Coherent Signal-subspace Method,CSM)两种宽带自适应波束形成方法。结合仿真实验,对这几个方法在波束形状、抑制干扰信号能力、运算速度等方面的性能进行比较。
[Abstract]:Beamforming is one of the core technologies of microphone array signal processing. Most of the traditional beamforming methods are based on the far-field narrow-band signal processing, which is suitable for the target signal in the far-field range of the array. When the microphone array system is used in the near field environment, such as small conference room, hands-free telephone and so on, if the far-field signal processing model is used, the effect of signal processing will be obviously decreased. In addition, in the near-field application environment of microphone array, the target signal comes from wideband signal such as voice. Therefore, the study of near-field narrowband and wideband beamforming methods for microphone array is more suitable for the practical application of microphone array system. In this paper, the method of microphone array beamforming based on near-field spherical wave model is studied. First, the near-field signal processing model is studied. According to the near-field signal propagation model, the near-field array response matrix is obtained, and the uniform linear array near-field guidance vector is obtained. Compared with the far field array response matrix, the near field array response matrix varies with azimuth and distance, while the far field array response matrix is only affected by the azimuth change. Secondly, near-field narrow-band beamforming methods are studied, including conventional beamforming, minimum variance Variance Distortionless response minimum Variance Distortionless response beamforming and least mean square error least Mean squared LMS algorithm. Recursive Least Mean Square-RLSs (recursive least squares) algorithm and sample Matrix inversion (Smi) algorithm. These narrowband beamforming methods are simulated based on the near-field uniform linear array. The performance of these methods in near-field environment such as resolution, anti-jamming ability, operation speed and ability to suppress interference signals are compared. Finally, the near-field broadband beamforming method is studied. Aiming at the particularity of wideband signal processing, three kinds of beamforming methods are introduced, including conventional time-delay beamforming based on time-domain signal processing. Incoherent Signal-subspace method (ISM) and coherent Signal-subspace method (CSM) are two kinds of wideband adaptive beamforming methods based on frequency domain decomposition. Combined with simulation experiments, the performance of these methods in beam shape, interference suppression ability and operation speed are compared.
【学位授予单位】：内蒙古大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN912.3

【相似文献】