【作者】 王德俊；

【导师】 惠俊英；

【作者基本信息】 哈尔滨工程大学 ， 水声工程， 2004， 博士

【摘要】 声传播和信号处理模型是水声系统的两个基础理论,基于矢量水听器的水声系统也是如此。矢量水听器信号处理技术经过十几年快速发展,展示出相比于声压信号处理技术的优点后,它在物理和信号理论方面需要更深入的基础研究,以更明确、更好地挖掘其内在的潜力。本文尝试在这两方面做一些基础性的探索研究。 “智能”水雷声引信、目标低噪声辐射声场测量和港口目标声特性监测等场合均关注近程(2Km以内)声场。因而,本文探讨近程声场的预报方法。 侧面波对近程声场有重要贡献,已发表的公开文献中仅对侧面波声场作近似估计,本文指出这些对侧面波声场的预报误差均较大。本文给出了侧面波声场的精确数值预报方法,从而有可能较精确预报近程声场。 只有在近程声压场精确预报的基础上才能用数值差分法计算振速场,本文的近程声场预报方法符合此要求。 由于矢量传感器正逐步被推广应用,对近程矢量场增加了解是必须的。本文通过数值计算,初步描绘了振速场的物理图像。 本文对比了波动理论、简正波声场和射线模型的声场预报结果,给出了后二种方法的近似程度。仅考虑简正波声场仅适合远程声场预报,它不适合近程声场预报。射线模型适合高频声场的预报,当kh>100时射线模型声场预报有足够高的近似精度。 本文另一部分内容讨论矢量信号处理的基础理论。基于声能流的方法,解决了声压信号处理固有的部分缺陷,提高了处理增益和方位估计精度,试验证明是比较实用的方法。对这种方法作理论建模是必要的,是有实际意义的。 本文推导了矢量水听器频域测量模型和模型参数。该测量模型便于推导互谱、平均声强以及方位估计的概率密度,进而深入分析矢量信号处理方法的性能。 根据频域测量模型,推演得到互谱、平均声强等声能流的概率密度公式,更多还原

【Abstract】 Sound propagation and signal processing model are the two fundamental theories of underwater acoustic systems, which are the same to that based on vector hydrophone. When vector sensor processing techniques have shown many advantages than pressure after more than ten years of rapid development, it is necessary to do more fundermental researches on physics and signal theory to explore the potential of vector hydrophone more specifically and effectively. We attempt to make probe fundermentally in above two espects in the thesis.Near distance acoustic fields are payed attention in intelligent acoustic mine and the measurement of tranquil targets’ radiate field and acoustic feature monitor of targets in the haven, etc. So in the thesis methods of acoustic field prognosis of near distance are discussed.Lateral wave contributes greatly to near distance acoustic field, which is only estimated approximatively in open literature and these have great errors. In this thesis accurate numerical prognosis method for lateral wave field is given, so the near distance acoustic field may be predicted accurately.Only based on accurate prediction of near distance pressure field the particle volicty field can be caculated by numerical differential method. This is satisfied by the near distance acoustic field method studied in this thesis.Since vector hydrophone is being widely applied step by step, it is necessary to comprehend near distance vector field. By numerical analysis physical concepts about the particle velocity field are descripted in the thesis.Results of acoustic field prognosis on wave theory, normal model and ray model are compared, and the approximate degree of the later two models is given. Normal mode is fit to forecast far distance acoustic fields only, but not near distance. Ray model is fit to forecast high frequence acoustic fields. If kh>100, ray model is accurate adequately.In the other part of the thesis the fundamental theories of vector signal processing are studied. Methods based on acoustic energy flow solve someinherent shortcomings of traditional pressure signal processing methods and increase processing gain and DOA estimation accuracy. Methods based on acoustic energy flow have been verified to be practically useful by trials. And to model those methods is necessary and practically significant.The frequency measurement model and its parameters are deduced in the thesis. Based on this model it’s convenient to express the probability densities of cross spectrum and average acoustic intensity and DOA, and to make deep analysis of vector signal processing methods’ performance sequentially.Based on the frequency measurement model, probability densities of acoustic energy flows such as cross spectrum and average acoustic intensity are deduced, then ROC (receiver operating characteristic) and vector array’s gain are theoretically analysed. In isotropic noise field, acoustic energy flows’ ROC is 4dB higher than pressure’s. Vector array’s gain is higher than that of pressure array. Under the condition of high snr, the vector array’s gain based on high orders surpasses that based on acoustic energy flows.Based on frequency measurement model probability densities and Crame-Rao Low Boundary (CRLB) of DOA estimation are deduced. At low snr DOA estimation is bias. If snr is higher than 7dB/Hz, the estimation is an unbiased MLE (Maximum Likelihood Estimation) and its accuracy approaches CRLB.The studies of near distance acoustic field and signal processing theories coincide with simulations and sea (lake) trials rationally.Exordium of the thesis introduces fundamental knowledge about theories and techniques of vector hydrophone, and overviews the corresponding development status at home and abroad, so that readers can have a general acquaintance with them.更多还原