【作者】 李婷婷；

【导师】 李秀坤；

【作者基本信息】 哈尔滨工程大学 ， 信号与信息处理， 2013， 博士

【摘要】 由于受各种复杂海洋环境因素的影响，水下目标探测与识别仍是水声信号处理中的重要难题之一。如何在复杂的海洋环境下获取高信混比(信噪比)的目标回波数据是提高水下目标探测能力与识别精度所迫切需要解决的关键问题。本文围绕水雷目标探测问题，对抑制主动声纳的混响干扰、获取高信混比的目标回波这一问题展开研究。本文研究的思路是基于基阵信号处理手段，利用小尺度基阵，通过形成空间的窄波束来实现目标方位估计的同时降低混响干扰。再依据目标回波和混响在信号特性上的差异，利用信号处理手段滤除混响。通过在空间域和变换域上实现混响干扰的二次抑制，以提高目标回波信号的信混比。文中首先对主动声纳探测目标所涉及到的三类信号形式(目标回波、混响、海洋环境噪声)在信号模型和信号特性上的差异进行了分析和总结，给出抑制混响干扰的处理思路，为后续处理提供理论依据。为了兼顾波束形成高的空间分辨率和稳健的性能，利用四阶累积量对阵列孔径的扩展特性，结合最小冗余阵列的构造，给出一种基于最小冗余阵列的四阶累积量波束形成方法。该方法仅利用少的基元数便可以实现高的空间分辨率的同时使四阶累积量数据矩阵构造的复杂度由原来的M~2×M~2降低到M×(2M1)。文中着重推导了四阶累积量波束形成处理增益的理论值，证明了在特定噪声分布条件下的阵增益与常规波束形成的阵增益之间存在临界信噪比。同时对四阶累积量波束形成方法在分辨率、快拍数影响、误差等方面的性能进行了深入讨论。并结合实际数据，对四阶累积量的空间分辨率进行了验证，结果表明基于四阶累积量波束形成方法比常规波束形成方法具有更高的空间分辨率和背景干扰抑制能力。针对宽带探测系统的处理要求，文中结合分数阶Fourier变换将基于四阶累积量波束形成方法扩展到LFM信号的方位估计中，并进行了性能分析。同时，利用目标回波信号和混响在分数阶Fourier变换后的差异，对混响信号进行变换域下的滤波。文中对单LFM信号、亮点模型信号叠加实测混响的情况进行了仿真分析，结果表明分数阶Fourier变换可以有效地滤除混响干扰，进一步提高目标回波信号的信混比。以分数阶Fourier变换为纽带，文中还给出了一种结合波束形成和变换域抗混响的联合处理方法。最后利用主动声纳海上目标探测数据进行了文中方法的验证。文中分别给出了海上实验数据单脉冲下的目标回波空间谱和在特定角度下的数据走航匹配结果。并对实测混响的概率密度进行了分析和拟合，给出了实测混响背景下四阶累积量波束形成的理论处理增益。最后对主波束回波信号进行了分数阶Fourier变换域上的混响滤波处理，进一步提高了目标回波的信混比。以文中的信混比定义方式，定量的给出了在两种掠射角下的平均处理增益。论文针对主动声纳水下目标探测的需求，实现了小尺度基阵的窄波束探测。相关研究方法可以简化水声探测设备和信号处理复杂度，并可以广泛应用于其它类型的水下目标探测中。更多还原

【Abstract】 Due to all kinds of complicated influences of sea environment, how to detect andrecognize underwater target is still one of the essentials in underwater acoustic signalprocessing. How to obtain target echo data with high SRR (Signal-to-Reverberation Ratio)(SNR：Signal-to-Noise Ratio) is the key issue to improve the ability of underwater targetsdetection and the accuracy of recognition in complex environments. Methods to suppressreverberation interference of active sonar and obtain target echo with high SRR data havebeen proposed in this paper, which are typical underwater mines detection problems.Firstly, array signal processing method, which is used to form narrow beams in spacedomain by small size arrays, realizes both target DOA (direction of arrival) estimation andreverberation suppression. Simultaneously, basing on the characteristic differences betweentarget echo and reverberation, a signal processing method is taken to filter reverberation.Adopting twice suppression of reverberation interference in both spatial domain andtransform domain, the purpose to improve the target echo SRR can be achieved.The differences on models and characteristics of three types of signals (the target echo,reverberation, noise) related to active sonar target detection are analyzed and summarized inthis paper. And a suppression scheme of reverberation interference is provided which is also atheoretical basis for the subsequent processing.In order to balance high spatial beamforming resolution and robust performance, thefourth-order cumulants which has the array aperture extension characteristics is used.Combined with minimum redundancy array structure, a fourth-order cumulants beamformingmethod based on minimum redundancy array is proposed. The method uses only a few arrayelements to achieve high spatial resolution, and reduces matrix complexity of data matrix ofthe fourth-order cumulants from the original M~2×M~2to M×(2M1). The theoreticalgain of fourth-order cumulants beamforming processing is calculated in this paper, whichproved that under some special noise distribution, there is a critical SNR between the gain offourth-order cumulants beamforming and that of conventional beamforming. At the same time,the performances of the fourth-order cumulants beamforming are fully discussed, includingresolution, snapshot influence, error performances. In order to verify the spatial resolution ofthe fourth-order cumulants beamforming, the experiment data from lake is processed, resultsof which proved the fourth-order cumulants beamforming method, which has higher spatialresolution and better background suppression ability, is better than the conventional beamforming method.To fulfill the requirements of broadband detection, the fourth-order cumulantsbeamforming combined with the fractional Fourier transform is extended to realize DOAestimation of LFM signal, and the performance of the method is also analyzed. At the sametime, based on the aggregation differences between target echo and reverberation afterfractional Fourier transform, the reverberation can be filtered in transform domain. The singleLFM signal, highlight model signal added with real reverberation signal are processed, andthe results showed that the reverberation interference can be filtered by the fractional Fouriertransform effectively, and the SRR of the target echo can be further raised. Taking fractionalFourier transform as a link, method combined target DOA estimation with anti-reverberationis given in this paper.The active sonar sea target detection data is dealt with to validate the characteristics ofthe proposed method finally. Both the target DOA estimation of single pulse and sailing matchresults in particular angle are presented in the paper. The probability density of realreverberation is analyzed and fitted. Moreover, under given reverberation background, thefourth-order cumulants beamforming processing gain is calculated.Then the fractional Fouriertransform domain reverberation filtering method is adopted for the main beam output signal.Based on the SRR definition in the paper, the averages of SRR processing gain under twograzing angles are obtained.According to target detection demands of underwater active sonar, a method by usingonly a small apture of array obtaining narrow beamwith is proposed. Related researches inthis paper can simplify the complexity of acoustic detection and signal processing, and can bewidely applied to other types of underwater targets detection.更多还原