【作者】 孙国仓；

【导师】 惠俊英；

【作者基本信息】 哈尔滨工程大学 ， 水声工程， 2008， 博士

【摘要】 水声物理、信号处理和海洋环境的紧密结合是声纳技术发展的必然趋势。浅海低频声纳是近代声纳发展的方向,水声探测的频段越来越低,矢量传感器的作用得到重视。在此背景下,本文对浅海矢量声场及其信号处理作了研究。全文的主要研究内容有以下5项,前2项为垂直声强信息的研究,后3项为水平声强信息的研究：(1)探讨了Pekeris波导中声压场和质点振速场的联合描述,尤其关注垂直声强流的分析。研究表明,由于各阶简正波的干涉作用,水平和垂直声强流均既有有功分量,亦有无功分量。低频声场的垂直声强流的无功分量虽不参与声能的输运,当单个声矢量传感器适当放置后却可用以判断声源的特定深度。这对于矢量信号处理是有意义的。给出了声压和质点振速联合互谱处理进行目标深度判别的算法。(2)进一步地研究了更为广泛的浅海波导中的垂直声强流。结果表明,所提出的利用垂直声强流无功分量判断目标的特定深度这一原理对不同浅海环境是可以适用的。近似的理论分析为传感器布放深度提供了指导,指出了所研究条件下垂直声强流的有功分量的符号分布与传感器布放深度无关。应用矢量声场的互易原理对不同浅海环境的矢量声场主要是垂直声强流无功分量进行了计算。在近距离处声场表现出近场特性,稍远距离处垂直声强流的符号即表现出规律性；海洋声速剖面对所提出原理的适用性影响不大；在规定作用距离后,海底吸收存在时对原理的应用不构成障碍,在一定程度上反倒有利；不同频率的声场的干涉图案有细微差别,但是传感器的布放位置可以是固定的,换言之,单个矢量传感器即可工作于较宽的频段。导出了传感器布放深度公式。这些结论以及所提出的原理是有条件的：波导中只存在2阶波导简正波。这在甚低频才可以做到,而近代声纳的工作频段的发展趋势正是如此。(3)针对矢量阵高分辨算法的运算量大的特点,尝试将声压阵中行之有效的减小计算量的方法推广到矢量阵中。提出了广义转换矩阵,使得矢量酉MUSIC算法得以成立。由于该算法利用了观测数据的复共轭数据,使得特征分解可以在实数域进行,不但减小了运算量而且提高了方位估计性能。将传播算子算法推广到宽带矢量阵处理中,由于不需要进行自相关矩阵的特征分解,该算法运算快速。仿真和实验数据处理表明所提出的算法是有效的。(4)声矢量传感器能够同时拾取空间同一点的声压振速信息,可以进行声场声强测量。据此提出了阵列声强器的概念。即当声矢量传感器成阵时,在波束域进行声强处理,以获得目标的方位信息。当空间高度欠采样时,矢量阵将产生栅瓣模糊,利用阵列声强器获得的方位信息可以抑制稀疏阵的栅瓣模糊,使得矢量阵在不增加阵元个数的前提下增大了阵列孔径,从而提高了方位分辨力。对阵列声强器去模糊性能进行了理论分析,仿真结果表明了所提算法的有效性。(5)对声压和振速信息进行联合处理的思想在目标方位估计中已有成熟应用,在时频参数估计中进行联合处理尚不多见,本文提出了以声压振速互协方差矩阵为基础的ESPRIT频率估计算法,在不损失估计性能的前提下,计算量减小为现有算法的1/27。给出了所用模型的克拉美劳下限。仿真结果表明所提算法是有效的。更多还原

【Abstract】 One of the inevitable tendencies of sonar technology is physics of underwater acoustics, signal processing and ocean environment being combined tightly. The direction of modern sonar’s development is low frequency and working in shallow water. With respect to lower and lower frequencies, more attentions are paid to acoustic vector sensor. Under this context, the dissertation studies acoustic vector field and its signal processing. The main contributions are as follows, the first two sections of which concern vertical acoustic intensity, the last three sections of which concern horizontal acoustic intensity:(1) The combined descriptions of the pressure field and particle velocity field in Pekeris waveguide, especially the vertical acoustic intensity flux are proposed in this paper. The results of the study show that both the horizontal and the vertical acoustic intensity flux have active and reactive component because of the interference between the normal modes. When an acoustic vector sensor is placed appropriately, the reactive component of the vertical acoustic intensity flux in low frequency acoustic field can be used to tell the source’s specified depth, although it can’t transport energy. Then the reactive component of the vertical acoustic intensity flux is of importance for vector signal processing. The pressure and particle velocity cross spectra signal processing algorithm is proposed to distinguish the targets.(2) The further study on vertical acoustic intensity flux in broad sense shallow water waveguide is proposed. The results show that the theory of using reactive component of vertical acoustic intensity flux to tell the source’s specified depth is valid for varies shallow water environment. Approximate theory analysis provides instruction to the placement of vector sensor, and points out that the polarity distribution of active component of vertical acoustic intensity flux is independent of the vector sensor’s depth.Reciprocal relationship is applied to compute the acoustic vector field especially the reactive component of vertical acoustic intensity flux in shallow water. The acoustic field shows characteristics of near field in the near range, and the polarity of reactive component of vertical acoustic intensity flux varies regularly in the far range; the sound speed profile affects the method proposed here slightly; the attenuation of the bottom doesn’t effect the method, on the contrary, it is in favor of the method; the interference patterns are different slightly for different frequencies, but the depth of the sensor can be fitted, in other words, one single vector sensor can work in a wide band. The formula of placing depth is proposed. All this conclusions is under the conditions that there are only two trapped normal modes. The condition can be fulfilled only at low frequencies, and the frequency band used by modern sonar is heading this direction.(3) There are much more output channels when the acoustic vector sensor is formed into arrays, which make the computation intense. A few computational efficient algorithms are proposed to apply to acoustic vector sensor array.A generalized transform matrix is constructed, and then the unitary MUSIC algorithm is adapted to acoustic vector sensors array. The observation data are incorporated with their conjugate, and then the covariance matrix can be decomposed in real-valued space. Hence, the proposed algorithm has features of lower computational complexity and better performance. The propagator method without eigen decomposition is applied to wideband coherent sources. The simulations and lake trial data processing show that the new algorithms are valid.(4) The array intensity estimator was proposed on the basis that acoustic vector sensors measure the pressure and particle velocity information of one spatial point simultaneously. After extracting the available spatial spectrum in the beamspace, one can process the acoustic intensity information to get the DOA of the target. When the spatial sampling law can’t be fulfilled, the spatial spectrum will lead to cyclically ambiguous DOA estimates. Using the array intensity estimator can solve the ambiguity problem of a sparse acoustic vector sensor array, thus the aperture of the array can be extended without increase the number of the element and then offers enhanced spatial resolution. The performance of the array intensity estimator was analyzed. The simulation experiments show that the algorithm is valid.(5) Estimating the direction of arriving using the pressure and velocity combined processing are well documented in the literature, few attentions have been put into the time and frequency estimates with the combined processing idea. A new ESPRIT algorithm for frequency estimating is proposed based on the pressure and velocity cross covariance matrix, which can reduce the computational burden to 1/27 of the previous algorithm. The Cramer-Rao lower bound is developed for the data model. The simulation shows the validity of the new algorithm.更多还原