【作者】 陈韶华；

【导师】 相敬林；

【作者基本信息】 西北工业大学 ， 通信与信息系统， 2006， 博士

【摘要】 本文的海洋声源信息包括运动声源(如舰船、潜艇、无人航行器等)信息和海洋环境噪声信息。这些信息是研究各种水中探测与通信技术乃至研究海洋气候与海洋生态环境的重要理论依据和实验验证的依据。本文研究关于在小尺度平台上进行海洋声源信息的获取与传输的技术。小尺度平台，包括锚系平台和运动平台，由于适合大时空跨度的数据测量，近年来引起了国内外的广泛关注。在小尺度平台上获取并传输声源信息，其难点是在复杂的海洋环境噪声背景中对运动声源的低频信号检测、方位估计以及高速水声数据传输。本文研究内容主要包括舰船噪声与海洋环境噪声特性，强噪声背景下微弱信号的预警检测，声强向量检测，基于声强向量阵的目标定向定位与尺度估计，海洋声信道特性与信道多径时延估计，通信信道的稀疏均衡与分数间隔均衡，信道的盲均衡等。本文研究涉及的理论基础包括数字信号处理、信号检测与参数估计、声强向量理论、自适应信号处理、阵列及通信信号处理等。本文主要研究工作的新进展与创新点包括：1．提出了一种任意空间分布任意阵形噪声场的数值模拟方法，用于本文基阵设计及检测定位方法的计算机仿真，也可用于一般阵列和声场的仿真计算。2．根据奈曼—皮尔逊准则分析了能量(幅度)统计平均检测、过零检测和功率谱检测等三种预警检测技术的统计性能，给出了实测船噪声的仿真结果。3．提出了基于声强向量阵的检测和方位估计方法，以克服常规阵列处理方法对低频声源的检测定位需要的阵列尺度太大的困难。分析得出了该检测方法的处理增益，推导了声强向量的概率密度函数，得到了声强向量阵检测方法的统计性能，并与基于声压的检测方法作了比较。理论分析与半消声室实验均验证了声强向量检测方法相对于声压处理的优越性能。基于声强向量阵的空间指向性，研究了宽带相干干扰背景下线谱和宽带连续谱的检测方法，给出了理论分析和实现方案。分析了方向相反的相干干扰互相抵消的原理，并在实验数据处理中观察到了这一现象。4．研究了三轴四元非典型声强向量阵对低频声源全空间定向的方法，以克服三轴六元典型声强向量阵在小尺度平台上难以布设的困难。系统地分析了其定向误差，包括有限差分误差、通道失配误差和环境噪声引起的随机误差。理论分析和计算机仿真均表明，非典型阵可在0.2m的小尺度平台上实现对低频声源的定向，其定向精度在本文实验室条件下可达到2°。5．提出了声强向量法和时延估计法的基阵共用技术，用于对体积目标三部位定向定位与尺度估计。理论分析、计算机仿真及实验结果均表明，共用同一小尺度基阵可同时实现对舰船中部、中后部和尾部的定向和区分，并用于估计目标尺度。6．结合MODE与带惩罚函数的NLS方法，研究了水声信道高分辨多径时延估计问题。采用混合算法处理了实测海试数据，基于处理结果重建的接收信号波形与实验接收波形吻合较好，从而验证了混合算法的有效性。7．提出了基于自适应延迟滤波器的DSPCC算法，通过极性相关序贯估计多径时延，对于确定时延通过自适应滤波估计多径幅度。仿真结果表明，DSPCC算法也具有对多径时延的高分辨能力，且运算量小。最后用该算法处理了海试数据，结果表明它比传统方法有更好的分辨性能。8．提出了一种改进的DFE均衡器，用于解决长时延扩展稀疏信道的均衡问题。根据信道冲激响应的前达分量强度，预测反馈滤波器重要抽头的位置，只对这一小部分抽头系数进行迭代运算，因而运算量显著减小，收敛速度加快并因而改善了跟踪信道时变的能力，但均方误差性能几乎没有损失。9．提出了一种T／2分数间隔稀疏CFE结构，以避免接收机对于符号定时误差的敏感性，并有效利用长时延扩展多径信道的稀疏性来降低均衡器的复杂度。理论分析与基于实测信道的计算机仿真表明，T／2-SCFE均衡器对符号定时误差保持了稳健性，性能优于符号间隔CFE及分数间隔DFE。10．提出了一种基于常数模准则的盲均衡方法ACMA，及ACMA与DFE盲判决反馈的联合均衡方法。采用概率统计和误差性能表面方法分析了其收敛性能，分析结果表明ACMA方法优于通常的CMA2-2方法。对特征值弥散达192的三径信道的仿真结果表明，ACMA-DFE盲均衡算法的最小均方误差较CMA2-2方法约低7dB，有较高的应用价值。更多还原

【Abstract】 In this dissertation, ocean acoustic information implies the information of moving soundsources (ships, submarines, autonomous vehicles, etc) and the information of sea ambient noise.This information is essentially important to study the underwater signal detection andcommunication, and even to study ocean climate or acoustic environment, both theoretically andexperimentally.This dissertation is intended to study the key techniques on ocean acoustic informationacquisition and transmission with a small size platform. Small size platforms, including anchoredand moving platform, have drawn extensive interest all over the world, since they are suitable fordata acquisition of large space-time span. The main difficulties of information acquisition andtransmission with a small platform lie in low frequency weak signal detection, bearing estimationof moving sound sources in complicated sea ambient noise background, and high speed digitaltransmission in multi-path underwater acoustic channels.The main topics studied in this dissertation focus on characteristics of ship-radiated noise andsea ambient noise, weak signal detection, detection with acoustic intensity array, bearing and sizeestimation of sound sources, multi-path time delay estimation, sparse equalization, fractionallyspaced equalization, and blind equalization of underwater acoustic communication channels. Thetheoretical basis involved in this dissertation include digital signal processing, signal detectionand estimation, vector acoustic intensity theory, adaptive signal processing, array andcommunication signal processing.The main contributions of this dissertation are as follows:1. A numerical method for simulating noise field of arbitrary spatial distribution and ofarbitrary array shape is presented. This method is used for the analysis and validation ofperformance of array design, signal detection and bearing estimation in this dissertation.2. The statistical detection performance of energy detector, zero-crossing detector and powerspectrum detector is analyzed based on the Neyman-Pearson criterion. Theoretical analysis andsimulations on ship-radiated noise measured at sea show that zero-crossing detector has goodperformance.3. A new method of signal detection for low frequency sound sources with vector acousticintensity array (VAIA) of limited aperture is proposed. The spatial-time gain for acoustic signaldetection in isotropic noise is derived, and the ROC (Receiver Operating Characteristic) inGaussian noise is obtained. Both theoretical analysis and experimental results show that acousticintensity detection is superior to sound pressure detection.4. A new method of two-dimensional DOA (Direction of Arrival) estimation for low frequencysound source with the non-typical VAIA of small size is proposed. The azimuth and pitch anglesof the sound source can be estimated with the 3 intensity components measured with this array.The estimation errors, including the finite difference approximation error (FDAE), theinstrumentation channel mismatch error (ICME), and error caused by ambient noise, are studiedsystematically. Experiments conducted in semi-anechoic chamber show that, after correcting theFDAE and ICME, the precision of DOA estimation in isotropic ambient noise is about 2°.5. A novel scheme is proposed for the bearing estimation of the separate sections of a volumetarget with a non-typical VAIA. The bearings of the midway section and the section between themidway and the stern of the volume target radiating low frequency noise is estimated with the vector acoustic intensity, while the bearing of the stem section radiating high frequency noise isestimated with the high precision adaptive FIR time delay estimation method. Experimentsconducted in a semi-anechoic chamber show that bearings of the 3 sections can be estimated withsatisfying precision and the 3 sections are distinguishable. This result provides a new approachfor ship size estimation at short distance.6. A hybrid algorithm for the high resolution multi-path time delay estimation is presented.First, the estimate of multi-path time delay is obtained by using eigenvalue decomposition withthe MODE algorithm, then the time delay and the corresponding gain estimate is refined with theNLS (nonlinear least square) estimator. Simulation on the linear frequency modulated (LFM)signal measured at sea is presented to prove the effectiveness of our method.7. For the purpose of reducing the complexity of adaptive delay filter, a polarity correlator isadopted to estimate the time delay, thus avoiding direct estimation of the mean square errorfunction needed by the previously proposed MDRL (Maximum Deviation from Regression Line)algorithm, therefore resulting in much lighter computational load at little cost in performance.Computer simulations on the simulated underwater acoustic channel and data processing ofchannel measured at sea prove that our method is effective.8. A modified adaptive decision feedback equalizer (DFE) is proposed to exploit fully thelong sparse channels in high-speed underwater acoustic digital communications. Based on therelationships between the optimal feedforward filter, feedback filter (FBF) on the one hand andthe channel impulse response on the other hand, influence of the precursor with different strengthon the FBF is analyzed. Our modified DFE only works on the significant taps predicted with theabove analysis. Simulations on sparse channel measured at sea confirm that our modified DFEhas the advantages of lighter computational load and faster convergence over the conventionalDFE with only a very small cost in performance.9. A Y/2 fractionally-spaced sparse complete feedback equalizer (CFE) scheme is proposed toavoid the performance deterioration caused by symbol timing error in the receiver. Theperformance of the scheme is analyzed based on minimum mean square error (MMSE) criterion,and the MMSE performance is proved to be the same as that of the DFE. The scheme alsoeffectively exploits the sparseness of the multi-path channels with large delay spread and reducesthe complexity of the equalizer with small performance degradation. Theoretical analysis andcomputer simulations over field measured wireless channel and underwater acoustic channelshow that the scheme performs better than symbol-spaced CFE and T/2 fractionally-spaced DFE.10. A new blind equalization method named ACMA (Absolute Constant Modulus Algorithm) isproposed. We give a fairly detailed theoretical analysis based on EPS (Error Performance Surface)that proves that our ACMA method has the advantage of much faster convergence rate. To furtherreduce the mean square error (MSE), the kurtosis of the output signal of the equalizer iscomputed and compared with a pre-selected threshold. If the kurtosis is smaller than thethreshold, which means the eye-diagram is open, the equalizer turns to the DFE mode, so as toachieve much lower noise floor. Both theoretical analysis and computer simulations showpreliminarily that our method has much better performance than CMA2-2.更多还原