低成本无线声阵列网络的实时高效DOA估计研究

发布时间：2018-05-20 07:34

本文选题：无线声传感器阵列网络 + 波达角估计　；参考：《浙江大学》2016年博士论文

【摘要】：无线声传感器阵列网络(WASAN)将无线声传感器网络与传感器阵列处理相结合,因此兼具他们各自拥有的高精度、自组网、强隐蔽性等优势,广泛应用于低空无人机、地面重型车辆以及水下潜艇等国防安全领域的被动目标定位和追踪。然而受本地计算、通信带宽和能量供给方面的限制,低成本无线阵列实时高效DOA估计是一个极具挑战性的课题。近年来新兴的压缩感知技术,面向信号的稀疏特性,通过少量的随机观测数据高概率地恢复出原始信号,在本地资源受限和通信受限的WASAN系统中具有非常大的潜在应用价值。本文在综述了国内外关于WASAN被动目标定位最新研究进展的基础上,针对WASAN的资源受限问题从低功耗信号采集、高效实时数据传输、信息信号处理以及快速算法实现方面展开较为系统和深入的研究。本文的主要内容如下:1.提出了一种基于声阵列信号疏特性的随机压缩采样与重构方法。该方法基于阵列信号的频域稀疏特性,针对传感器节点本地计算资源和通信带宽受限的特点通过随机非均匀采样技术对原始信号进行降维观测,在降低WASAN数据发送量的同时无需额外的本地计算开销。2.提出了一种基于联合稀疏特性的DOA(DirectionofArrival)估计方法。该方法利用有限数目的目标在角度空间的稀疏性和声源信号在频域的稀疏性,通过引入了联合稀疏表示矩阵的方法对阵列随机压缩采样数据在角度空间和频率空间进行联合稀疏表示,并直接对目标的DOA角度进行稀疏重构。该方法在降低阵列数据量的同时避免了传统压缩采样-信号重构-DOA处理模式中信号重构误差到DOA估计误差的传递过程,减少了信号处理的冗余。3.提出了一种极低数据量的DOA估计方法。该方法面向DOA估计的本质问题-相位差估计,利用单比特压缩感知技术对信号波形进行重构的特点,在传感器节点端仅用1比特对采样得到的数据进行量化。进一步利用信号的联合稀疏特性,从单比特量测中估计出目标的DOA空间谱。4.提出了一种快速三维DOA估计方法用于WASAN系统本地DOA估计实现。该方法基于阵列几何结构特性解决了当阵元间距大于信号半波长时候存在的相位模糊问题并构建了基于相位量测的线性DOA估计模型直接在相位空间对DOA值进行计算。该方法解决了传统的AML、MUISC、Beamforming等基于角度穷举搜索算法计算复杂度大的问题。理论分析表明,该方法的DOA估计误差的协方差和AML估计器的克拉美罗下界一致。5.提出了一种基于相位量测的分布式快速目标定位算法。该方法仅需要将少量的本地预处理结果发送到后端融合中心进行直接定位,避免了传统基于完整信息直接定位方法的完整数据发送要求。理论分析表明该方法的估计误差协方差与传统的基于完整信息直接定位方法的克拉美罗下界一致。
[Abstract]:WASAN (Wireless Acoustic Sensor Array Network) combines wireless acoustic sensor network with sensor array processing, so they have the advantages of high precision, self-organizing network and strong concealment, so they are widely used in low-altitude UAVs. Passive target location and tracking in defense and security areas such as ground heavy vehicles and underwater submarines. However, due to the limitations of local computing, communication bandwidth and energy supply, real-time and efficient DOA estimation for low-cost wireless arrays is a challenging task. In recent years, the new compression sensing technology, which is oriented to the sparse characteristic of signal, recovers the original signal with high probability through a small amount of random observation data. It has great potential application value in WASAN systems with limited local resources and communication constraints. On the basis of summarizing the latest research progress of WASAN passive target localization at home and abroad, this paper aims at the problem of resource limitation of WASAN from low-power signal acquisition and efficient real-time data transmission. Information signal processing and fast algorithm implementation are studied systematically and deeply. The main contents of this paper are as follows: 1. A random compression sampling and reconstruction method based on the sparse characteristics of acoustic array signals is proposed. This method is based on the sparse characteristic of array signal in frequency domain, aiming at the local computing resource and the limited communication bandwidth of sensor node, the dimension reduction observation of the original signal is carried out by random non-uniform sampling technique. Reduce the amount of WASAN data sent without additional local computing overhead. 2. 2. A method for estimating load Direction of Arrivalbased on joint sparsity is proposed. This method utilizes the sparsity of a finite number of targets in angle space and the sparsity of sound source signals in frequency domain. The joint sparse representation matrix is introduced to represent the array randomly compressed sampling data in angle space and frequency space, and the DOA angle of the target is reconstructed directly. This method not only reduces the amount of array data, but also avoids the transfer process from signal reconstruction error to DOA estimation error in the traditional compression sample-signal refactor-DOA processing mode, and reduces the redundancy of signal processing. A very low data DOA estimation method is proposed. This method aims at the essential problem of DOA estimation-phase difference estimation. It uses single bit compression sensing technology to reconstruct the signal waveform and quantifies the sampled data only with 1 bit at the sensor node. Furthermore, the DOA spatial spectrum. 4. 4 of the target is estimated from the single bit measurement by using the joint sparsity property of the signal. A fast 3D DOA estimation method for local DOA estimation in WASAN systems is proposed. The method solves the phase ambiguity problem when the array spacing is larger than the half-wavelength of the signal based on the geometric structure of the array and constructs a linear DOA estimation model based on phase measurement to calculate the DOA value directly in the phase space. This method solves the problem of high computational complexity of traditional AML-MUISC-Beamforming algorithm based on angle exhaustive search. Theoretical analysis shows that the covariance of the DOA estimation error of this method is consistent with that of the lower bound of the AML estimator. A distributed fast target location algorithm based on phase measurement is proposed. This method only needs to send a small amount of local preprocessing results to the back end fusion center for direct location, thus avoiding the requirement of complete data transmission based on traditional direct localization method based on complete information. Theoretical analysis shows that the estimation error covariance of this method is consistent with that of Clemero lower bound which is based on the traditional direct location method based on complete information.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP212.9;TN911.7

【参考文献】