【作者】 张瑶；

【导师】 梁国龙；

【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】 为了提高主动声呐系统对目标的探测能力，本文以浅海水声环境为应用场合，以线谱信号为主线，分别针对高分辨能力声呐信号设计及其检测方法、线谱信号高分辨频率估计方法以及基于单频水声信号的多径时延估计算法等方面，进行了深入系统的理论和实验研究，主要研究内容包括：1.在对常用声呐信号以及传统梳状谱信号进行分析的基础上，提出了一种新的梳状谱信号设计方法。该波形设计方法通过在频点间加入一比例序列作为抗速度模糊频率间隔，有效的降低了多普勒频谱混叠，并根据梳状谱信号的频谱特点，利用分集发射的形式，并对每帧信号设定不同的多普勒测速范围及频带范围，不但更高效的利用频带资源，而且可以有效提高系统的探测帧率。仿真分析证明了本文提出的设计方法具有更好的时延与频移分辨能力，并有效提高了系统的测速精度以及混响抑制能力。针对梳状谱信号峰均功率比较高的特点，提出了一种改进的快速相位优化算法，新方法在保证计算量显著减小的前提下，能有效的降低发射信号的峰均功率比。2.针对梳状谱信号检测的问题，提出一种基于频域匹配搜索的梳状谱信号检测方法。梳状谱信号的多普勒容限较小，针对传统的多匹配/拷贝相关检测方法运算量较大的问题，本文根据梳状谱频谱多峰结构的特点，提出了一种基于频域匹配搜索的检测算法：利用信号与噪声频域分布的差别，通过在频域对频点匹配搜索积累的方法实现了对于多普勒回波信号的检测以及速度估计，有效的降低了运算量。针对低信噪比条件下信号检测的问题，根据梳状谱信号的特点，通过瞬时相关积分法有效提高了接收信号的信噪比。仿真实验证明在噪声、混响条件下，本文算法均具有较好的信号检测以及速度估计能力，将此预处理方法与频域搜索方法相结合可进一步提高其检测与估计性能。3.针对低信噪比条件下的线谱信号频率提取问题，提出了基于双递归自适应滤波的高精度频率估计算法。对线谱信号频率的提取一直是目标特征估计的重要参数，传统的频率估计方法如：短时傅里叶变换方法、准正交采样频率估计方法、过零频率估计方法、基于时频分析的频率估计方法等，或受限于计算量或对噪声比较敏感。基于二阶自适应陷波滤波器结构的自适应瞬时频率估计器，其算法简单，可实现对信号的滤波、幅度及相位估计，本文对其基本原理以及其性能进行了详细的描述，并推导了Notch滤波器滤波输出瞬时频率方差与滤波器中心频点以及信噪比条件的关系，根据以上分析提出了一种基于双递归自适应滤波的频率估计方法，通过将多个自适应Notch滤波器组成递归环路，通过对滤波器输出瞬时频率方差的分析进行滤波器参数的更新，以达到更优的滤波输出结果。仿真分析表明：双递归自适应滤波方法具有良好的噪声抑制能力，能够实现低信噪比条件下的信号检测与频率估计。4.针对线谱信号多途时延提取的问题，提出了基于同态滤波的时延估计算法。针对窄带信号多途时延估计的问题，推导了倒谱算法获取多途时延估计的原理，针对其对噪声较敏感的问题，提出了一种基于对数域谱减法的同态滤波时延估计算法：利用主动系统的特点，充分利用信号的先验信息，将信号项在对数域消除，并对相减后的信号通过滤波算法消除其残余的信号成分以及噪声项，再将信号恢复到时域，以得到多途时延信息。仿真实验证明，基于同态滤波的多途时延估计方法具有很高的估计精度，与直接进行复倒谱运算相比，有效的提高了其抗噪声能力，而且对于信号的多普勒频移也具有很好的适应能力。更多还原

【Abstract】 To improve the passive sonar system’s ability of target detection and parameterestimation, the shallow water environment was considered in this paper, against theline-spectrum signal, design and detection method of high resolution sonar signal, linespectrum signal high-resolution frequency estimation method and multipath time-delayestimation method for single frequency acoustic signal were analyzed researched theoreticallyand experimentally, respectively. The key contributions are as follows:1. Based on the analysis of common sonar signal and traditional comb spectrum signal, anew method of comb spectrum signal design was proposed. The Doppler spectrum aliasingwas reduced effectively by adding a ratio of equal sequence in frequency as an anti-speedfuzzy frequency interval. According to the spectrum characteristics of comb spectrum signals,a set of comb spectrum signals with different Doppler tolerance and frequency range werelaunched. Consequently, the efficiency of frequency resources and frame rate of the systemwere improved. Simulation results demonstrated the better time delay and frequencyresolution, as well as the improved speed measuring precision and reverberation suppressionability of the system. Focusing on the high peak to average power ratio (PAPR) of the combspectrum signals, a quick phase optimization algorithm was presented. The PAPR of thesignal was reduced effectively on the basis of dramatically decreased computation.2. A detection method of comb spectrum signal based on the matching search infrequency domain was proposed. The Doppler tolerance of comb spectrum signal is small.Considering the huge computation of traditional matching/copies correlation method andmulti-peak characteristics of comb spectrum signal, a detection algorithm based on frequencydomain matching search was proposed. The frequency distribution difference between signaland noise was used. Doppler echo signal detection and velocity estimation were achieved bysearching the matched point and accumulation in the frequency domain. Furthermore, theamount of computation was reduced effectively. In view of the signal detection problems inlow SNR, the signal-to-noise ratio of the received signal was greatly improved through theinstantaneous correlation integral method based on the characteristics of the comb spectrumsignal. Simulation results showed that in the noise and reverberation conditions, the proposedalgorithm had better performance of signal detection and velocity estimation. Theperformance could be further improved by combining the frequency domain search methodand instantaneous correlation integral method. 3. Focused on the frequency extracting in low signal-to-noise ratio, a high-accuracyfrequency estimation algorithm based on the double recursive adaptive filtering was proposed.The extracted line spectrum frequency was an important parameter to estimate the targetfeatures. The traditional frequency estimation methods, such as short time Fourier transformmethod, quasi-orthogonal sampling frequency estimation method, zero frequency estimationmethod and time-frequency analysis method, were either limited by calculation or sensitive tonoise. The algorithm of the frequency estimator which was based on second-order adaptivenotch filter (ANF) was simple. The signal filtering, amplitude and phase estimation could alsobe achieved. The basic principle and performance were described in detail. Based on theanalysis of the relationship of the output instantaneous frequency variance and centerfrequency of the notch filter versus SNR were deduced, a frequency estimation method usingthe dual-recursive adaptive filtering was presented. Multiple adaptive notch filters werecomposed as a recursive loop. A better filtering result was gained by updating the filterparameters based on the analysis of the output instantaneous frequency variance. Simulationanalysis showed that this method had good noise suppression capability and could achievesignal detection and frequency estimation in low SNR.4. To estimate the multipath timedelay, a time delay estimation algorithm based onhomomorphic filtering was proposed. Focusing on the narrowband signal multipathtime-delay estimation, the principle of obtaining multipath time-delay through cepstrumalgorithm was deduced. Concerning the problem of its sensitivity to noise, a homomorphicfiltering delay estimation algorithm based on spectral subtraction in logarithm domain wasproposed. Taking advantage of the active system and the priori information of signal, signalwas eliminated in the logarithmic domain. The subtracted signal was filtered to eliminate theresidual signal component and noise. Then the signal was recovered to the time domain to getthe multipath time-delay information. Simulation results showed that the multipath time-delayestimation method based on homomorphic filtering had high estimation accuracy. Comparedto the direct complex cepstrum method, the proposed method effectively improved theanti-noise ability and adapted better to the Doppler frequency shift of the signal.更多还原