CN103513249A

CN103513249A - Broadband coherent mold base signal processing method and system

Info

Publication number: CN103513249A
Application number: CN201210210580.8A
Authority: CN
Inventors: 巩玉振
Original assignee: Institute of Acoustics CAS
Current assignee: Institute of Acoustics CAS
Priority date: 2012-06-20
Filing date: 2012-06-20
Publication date: 2014-01-15
Anticipated expiration: 2032-06-20
Also published as: CN103513249B

Abstract

The present invention proposes a broadband coherent mode-based signal processing method and system. The method includes: Step 101) Obtaining array reception data of B frequency points, performing beam normalization on the data of the full frequency band, and coherently forming comprehensive reception data vector; Step 102) Obtain the modeling vectors of B frequency points, and coherently form a comprehensive expected steering vector for the modeling vectors of the full frequency band; Step 103) Obtain the optimal Steering vector; Step 104) Obtain the final target positioning result according to the optimal steering vector. The step 102) is: traversing possible positions of the sound source, inputting the weighted vectors of B frequency points into the modeling tool, and accumulating the B weighted vectors to obtain a comprehensive expected steering vector. Step 103) includes: correlating the integrated received data vector and the integrated expected steering vector and traversing all possible sound source positions to obtain a decision function; the position of the maximum value of the decision function is the sound source position.

Description

A broadband coherent mode-based signal processing method and system

技术领域 technical field

本发明属于声纳数字信号处理领域，特别涉及一种宽带相干模基信号处理方法。The invention belongs to the field of sonar digital signal processing, in particular to a wideband coherent mode-based signal processing method.

背景技术 Background technique

在声纳系统中，根据工作原理的不同，分为主动声纳和被动声纳。被动声纳本身不发射声信号，只是被动地接受可疑目标的辐射噪声，以进行目标检测，继而进行目标声源的定位以及识别。In the sonar system, according to the different working principles, it is divided into active sonar and passive sonar. Passive sonar itself does not emit acoustic signals, but passively accepts the radiation noise of suspicious targets for target detection, and then locates and identifies the target sound source.

被动声纳声源定位需要空间、带宽等信息，以得到较低的旁瓣，并提高对环境失配的鲁棒性以及对噪声的抑制能力。海洋传播信道是频率变化的，信道的衰落与时散特性在本质上是由于海洋环境地声属性的频率依赖性引起的。对固定的海域，采用单频或窄带信号进行模基处理可能会导致分辨率低、稳健性差及唯一性等问题，而用足够宽的宽带信号可以得到更优的处理结果。Passive sonar sound source localization requires information such as space and bandwidth to obtain lower side lobes, and improve the robustness to environmental mismatch and the ability to suppress noise. The frequency of the ocean propagation channel varies, and the fading and time-dispersion characteristics of the channel are essentially caused by the frequency dependence of the geoacoustic properties of the ocean environment. For fixed sea areas, using single-frequency or narrow-band signals for model-based processing may lead to problems such as low resolution, poor robustness, and uniqueness, while using sufficiently wide-band signals can obtain better processing results.

宽带信号处理方法可以分为两种：一种是传统的非相干方法，另一种是相干方法。非相干的处理方法利用了一个频点内的空间相干性，但没有考虑频率之间的相干信息，多个频率点的模糊表面直接进行实数域平均。目前的处理算法大多采用这种空间相干处理、频率间非相干处理的方法。而相干处理方法则考虑了频率间的互相关信息，并选择与频率有关的加权系数进行加权平均，由于考虑到了频率间的相干性，可以更充分地利用水声传播特性，从而提高定位的准确度。目前，对宽带相干处理的研究是一个热门课题。。Broadband signal processing methods can be divided into two types: one is the traditional non-coherent method, and the other is the coherent method. The non-coherent processing method utilizes the spatial coherence within a frequency point, but does not consider the coherent information between frequencies, and the fuzzy surfaces of multiple frequency points are directly averaged in the real number domain. Most of the current processing algorithms adopt this method of spatial coherent processing and inter-frequency non-coherent processing. The coherent processing method considers the cross-correlation information between frequencies, and selects frequency-related weighting coefficients for weighted average. Since the coherence between frequencies is considered, the underwater acoustic propagation characteristics can be more fully utilized, thereby improving the accuracy of positioning. Spend. Currently, research on broadband coherent processing is a hot topic. .

最初的宽带相干模基处理算法是在时域实现的，在文献1“C.S.Clay.Optimum timedomain signal transmission and source localization in a waveguide.J.Acoust.Soc.Am.,81:660-664,1987”中，最早实现了时域的宽带相干匹配场处理，他提出用测量得到的脉冲响应与建模得到脉冲响应相匹配，以此来估计声源位置，他还将此方法扩展到多阵元的互相关处理。文献2“R.K.Brienzo,W.S.Hodgkiss.Broadband matched-field processing.J.Acoust.Soc.Am.,94:2821-2831,1993”使用该算法在试验中成功地定位了9km处的爆炸声源。但是这种算法需要预先知道声源的频谱，这在大多被动声纳的应用条件下是不可能的。在文献3“P.Hursky，M.B.Porter,M.Siderius.High frequency(8-16hz)model-based source localization.J.Acoust.Soc.Am.,115:3021-3032,2004”中，研究了高频段（1kHz以上）时域宽带MFP的性能，由于高频段对于参数扰动更加敏感，声源定位更加困难。他们首先假设声源波形已知，从而使用匹配滤波的方法得到信道的脉冲响应，与建模得到的脉冲响应相匹配，即可得到声源位置的估计，当波形未知时，他们将两个水听器接收到的信号做互相关处理，从而消除了声源波形对处理器的影响，文中还给出了声源深度和距离的跟踪结果。The original broadband coherent mode-based processing algorithm is implemented in the time domain, in the literature 1 "C.S.Clay.Optimum timedomain signal transmission and source localization in a waveguide.J.Acoust.Soc.Am.,81:660-664,1987" In , he realized the wideband coherent matching field processing in the time domain first. He proposed to use the measured impulse response to match the modeled impulse response to estimate the position of the sound source. He also extended this method to multi-element Cross-correlation processing. Document 2 "R.K.Brienzo, W.S.Hodgkiss. Broadband matched-field processing. J. Acoust. Soc. Am., 94:2821-2831, 1993" used this algorithm to successfully locate the explosion sound source at 9km in the experiment. But this algorithm needs to know the frequency spectrum of the sound source in advance, which is impossible in most passive sonar application conditions. In the literature 3 "P. Hursky, M.B.Porter, M. Siderius. High frequency (8-16hz) model-based source localization. J. Acoust. Soc. Am., 115:3021-3032, 2004", the high The performance of the time-domain broadband MFP in the frequency band (above 1kHz), because the high-frequency band is more sensitive to parameter disturbances, it is more difficult to locate the sound source. They first assume that the waveform of the sound source is known, and then use the matched filter method to obtain the impulse response of the channel, and match the impulse response obtained by modeling to obtain the estimation of the location of the sound source. When the waveform is unknown, they combine the two channels The signal received by the earphone is processed by cross-correlation, thereby eliminating the influence of the sound source waveform on the processor. The paper also gives the tracking results of the depth and distance of the sound source.

文献4“E.K.Westwood.Broadband matched-field source localization.J.Acoust.Soc.Am.,91(5):2777-2789,May 1992”在频域实现了宽带相干模基处理，与时域宽带相干处理的思想一致，在频域中，也是将测量得到的脉冲响应与建模得到的脉冲响应做相关处理。与时域的方法不同，处理器输出不再是时域的相关函数的最大值，而是频域互相关的相干累加。由于该处理器在每个频点做了相关处理，所以不需要知道声源的频谱信息。文献4指出，宽带相干处理器好于非相干处理器，在相干处理器中，去掉自相关的部分（即CSDM中的对角项）效果会更好，此外，他还讨论了阵孔径、信号带宽对处理器性能的影响，使用更多的阵元、更大的带宽、更大的孔径将会获得更好地定位结果。文中还给出了5000m深海的试验结果，使用200米锚底垂直阵，成功的将声源跟踪到了42km的距离。Document 4 "E.K.Westwood.Broadband matched-field source localization.J.Acoust.Soc.Am.,91(5):2777-2789,May 1992" realized broadband coherent mode-based processing in the frequency domain, which is coherent with time domain broadband The processing idea is the same. In the frequency domain, the measured impulse response is also correlated with the modeled impulse response. Different from the method in the time domain, the processor output is no longer the maximum value of the correlation function in the time domain, but the coherent accumulation of the cross-correlation in the frequency domain. Since the processor performs correlation processing at each frequency point, there is no need to know the spectral information of the sound source. Document 4 pointed out that broadband coherent processors are better than non-coherent processors. In coherent processors, the effect of removing the autocorrelation part (that is, the diagonal term in CSDM) will be better. In addition, he also discussed array aperture, signal The impact of bandwidth on processor performance, using more array elements, larger bandwidth, and larger aperture will get better positioning results. The paper also gives the test results in the 5000m deep sea, using the 200m anchored vertical array, successfully tracked the sound source to a distance of 42km.

文献5“Z.-H.Michalopoulou and M.B.Porter.Matched-field processing forbroad-band source localization.IEEE Journal of Oceanic engineering,21:384-392,1996”提出了一种直接匹配声场的宽带相干处理方法（阵元归一化宽带相干匹配处理器，简称MP算法），区别于之前的匹配互相关的方法，该算法将阵列接收到的各频率的声场累接起来，形成一个超级向量，由于频率间的相位差影响，直接的声场匹配效果很差，他们提出了一种归一化的方法，每个频率的向量都以第一个阵元为参考归一化，然后将各频率的向量累接起来，形成超级向量，以消除频率间相位差的影响，此方法可以直接用于MVDR等自适应算法。他们将这种宽带相干的方法用于试验数据，结果表明MVDR宽带相干处理器对声源距离和深度的跟踪有效率达到90%，而非相干的MVDR仅为10%。但是这种方法在低信噪比时，归一化的效果明显变差，从而不能消除声源的影响。Document 5 "Z.-H.Michalopoulou and M.B.Porter. Matched-field processing for broad-band source localization. IEEE Journal of Oceanic engineering, 21:384-392, 1996" proposes a broadband coherent processing method that directly matches the sound field ( Array Element Normalized Broadband Coherent Matching Processor, referred to as MP algorithm), which is different from the previous method of matching and cross-correlation, this algorithm accumulates the sound fields of each frequency received by the array to form a super vector, because the frequency between Due to the influence of phase difference, the direct sound field matching effect is very poor. They proposed a normalization method. The vectors of each frequency are normalized with the first array element as a reference, and then the vectors of each frequency are accumulated. , to form a super vector to eliminate the influence of the phase difference between frequencies, this method can be directly used in adaptive algorithms such as MVDR. They applied this broadband coherent method to experimental data, and the results showed that the MVDR broadband coherent processor can track the distance and depth of the sound source with an efficiency of 90%, while the non-coherent MVDR is only 10%. However, when this method has a low signal-to-noise ratio, the effect of normalization becomes significantly worse, so the influence of the sound source cannot be eliminated.

本发明的技术方案所采用的技术手段有效的克服了文献5存在的技术缺陷。现有技术的模基信号的处理方法如图1所示。The technical means adopted in the technical solution of the present invention effectively overcome the technical defects in Document 5. The processing method of the modulus signal in the prior art is shown in FIG. 1 .

发明内容 Contents of the invention

本发明目的在于，为克服现有宽带信息的非相干处理带来的处理损失，在低信噪比的情况下目标检测能力下降，提出一种宽带相干模基处理方法及系统。The purpose of the present invention is to propose a broadband coherent mode-based processing method and system in order to overcome the processing loss caused by the non-coherent processing of the existing broadband information, and the target detection ability declines in the case of low signal-to-noise ratio.

本发明所述的波束归一化宽带模基处理算法，将每个频点的阵列信号用波束形成后的参考向量归一化，从而降低处理器的最小可检测信噪比，提高检测性能。The beam-normalized broadband modulus processing algorithm of the present invention normalizes the array signal of each frequency point with the reference vector after beamforming, thereby reducing the minimum detectable signal-to-noise ratio of the processor and improving the detection performance.

为实现上述目的，本发明提供了一种宽带相干模基信号处理方法，所述方法包含：In order to achieve the above object, the present invention provides a method for processing a broadband coherent mode-based signal, the method comprising:

步骤101）获得B个频点的阵列接收数据，对全频带的数据做波束归一化，相干形成综合接收数据向量；Step 101) Obtain the array reception data of B frequency points, perform beam normalization on the data of the whole frequency band, and coherently form a comprehensive reception data vector;

步骤102）获得B个频点的建模向量，并对全频带的建模向量相干形成综合期望导向矢量；Step 102) Obtain the modeling vectors of B frequency points, and coherently form a comprehensive expected steering vector for the modeling vectors of the whole frequency band;

步骤103）依据所述的综合接收数据向量和综合期望导向矢量，得到最优导向矢量；Step 103) Obtain the optimal steering vector according to the integrated received data vector and the integrated expected steering vector;

步骤104）依据最优导向矢量获得最终的目标定位结果。Step 104) Obtain the final target positioning result according to the optimal steering vector.

上述技术方案中，所述步骤101）具体为：对接收的B个频点的阵列接收数据进行快速傅里叶变换，获得各个单频点的数据；对B个频点的数据依次进行波束形成和归一化处理，形成综合接收数据向量；其中，B取值为大于等于1。累积L个快拍的阵列接收数据构造阵列协方差矩阵。In the above technical solution, the step 101) is specifically: perform fast Fourier transform on the received array reception data of B frequency points to obtain the data of each single frequency point; perform beamforming on the data of B frequency points sequentially and normalization processing to form a comprehensive received data vector; wherein, the value of B is greater than or equal to 1. The array received data of L snapshots is accumulated to construct the array covariance matrix.

上述技术方案中，所述步骤101）进一步包含如下子步骤：In the above technical solution, the step 101) further includes the following sub-steps:

步骤101-1）用线阵接收空间信号，得到一个快拍时刻N个阵元的时域信号；Step 101-1) Receive spatial signals with a line array to obtain time domain signals of N array elements at a snapshot time;

步骤101-2）对时域数据做快速傅里叶变换，得到B个频点的阵列数据(X₁,X₂，...,X_B)，其中第i个频点的阵列数据Xi表示如下Step 101-2) Perform fast Fourier transform on the time-domain data to obtain the array data of B frequency points (X ₁ , X ₂ ,...,X _B ), where the array data Xi of the i-th frequency point represents as follows

X_i=[x_i1,x_i2,...,x_iN]^T X _i =[x _i1 ,x _i2 ,...,x _iN ] ^T

其中x_ij表示第j个阵元第i个频点的接收数据，T表示转置；Among them, x _ij represents the received data of the i-th frequency point of the j-th array element, and T represents the transpose;

步骤101-3）对B个频点的数据做波束形成，得到第i个频点的波束数据X_i，beam为Step 101-3) Perform beamforming on the data of B frequency points to obtain the beam data X _{i of the i-th frequency point, beam} is

X_i,beam=X_i*e_i X _i,beam =X _i *e _i

其中，e_i表示第i个频点的阵列方位补偿向量，由阵元位置、目标方位及信号频率决定；Among them, e _i represents the array orientation compensation vector of the i-th frequency point, which is determined by the array element position, target orientation and signal frequency;

将此波束数据作为每个频点的参考信号(X_1ref,X_2ref,...，X_Bref)，即Take this beam data as the reference signal of each frequency point (X _1ref , X _2ref ,..., X _Bref ), namely

X_iref=X_i,beam X _iref =X _i,beam

以此参考信号对每个频点的阵列数据(X₁,X₂,...,X_B)做归一化，得到归一化后的数据

如下式：Use this reference signal to normalize the array data (X ₁ ,X ₂ ,...,X _B ) of each frequency point to obtain the normalized data

as follows:

$(({\overset{^^}{X x}}_{11},, {\overset{^^}{X x}}_{22},, . . . . . .,, {\overset{^^}{X x}}_{B B})) = = ((\frac{{X x}_{11}}{{X x}_{11 ref ref}},, \frac{{X x}_{22}}{{X x}_{22 ref ref}},, . . . . . .,, \frac{{X x}_{B B}}{{X x}_{Bref Bref}}))$

步骤101-4）将B个频点归一化后的数据

累接起来，形成一个长度为B*N的综合接收数据向量

若记：Step 101-4) Normalized data of B frequency points

Tiered together to form a comprehensive received data vector of length B*N

If remember:

${\overset{^^}{X x}}_{11} = = {[[{\overset{^^}{x x}}_{1111},, {\overset{^^}{x x}}_{1212},, . . . . . .,, {\overset{^^}{x x}}_{11 N N}]]}^{T T}$

其中

表示第j个阵元第i个频点的归一化后的数据，则扩展后的归一化向量：in

Represents the normalized data of the i-th frequency point of the j-th array element, then the extended normalized vector:

${X x}_{e e 11} = = {[[{\overset{^^}{x x}}_{1111},, {\overset{^^}{x x}}_{1212},, . . . . . .,, {\overset{^^}{x x}}_{11 N N},, {\overset{^^}{x x}}_{21 twenty one},, {\overset{^^}{x x}}_{22 twenty two},, . . . . . .,, {\overset{^^}{x x}}_{22 N N},, . . . . . .,, {\overset{^^}{x x}}_{B B 11},, {\overset{^^}{x x}}_{B B 22},, . . . . . .,, {\overset{^^}{x x}}_{BN BN}]]}^{T T};;$

步骤101-5）接收L个快拍时刻的数据，分别按步骤101-2）-101-4）得到每个快拍的综合接收数据向量，全部L个快拍的扩展向量记为X_e，all；Step 101-5) Receive the data of L snapshot moments, and obtain the comprehensive received data vector of each snapshot according to steps 101-2)-101-4) respectively, and denote the extended vectors of all L snapshots as X _{e, all} ;

X_e，all=[X_e1,X_e2,..X_eL]X _{e, all} = [X _e1 ,X _e2 ,..X _eL ]

所述步骤101-1）是接收一个快拍时刻的数据，需要重复L次，步骤101-5）即是这个意思；The step 101-1) is to receive the data of a snapshot moment, which needs to be repeated L times, and the step 101-5) means this;

步骤101-6）使用L个快拍的综合接收数据向量形成阵列协方差矩阵Step 101-6) Use the integrated received data vectors of L snapshots to form an array covariance matrix

$R R = = E E. [[{X x}_{e e,, all all} {X x}_{e e,, all all}^{H h}]] - - - - - - ((11))$

其中，E表示做统计平均，H表示共轭转置。Among them, E means statistical average, and H means conjugate transpose.

上述技术方案中，所述步骤102）具体步骤为：遍历声源可能出现的位置，并结合海洋环境参数，输入建模工具得到B个频点的加权向量，将B个加权向量累接起来，得到综合期望导向矢量。In the above technical solution, the specific steps of the step 102) are: traversing the positions where the sound source may appear, combined with the marine environment parameters, inputting the weighted vectors of B frequency points into the modeling tool, and accumulating the B weighted vectors, Get the integrated desired steering vector.

上述技术方案中，所述步骤103）进一步包含：In the above technical solution, the step 103) further includes:

步骤103-1）将所述的综合接收数据向量和综合期望导向矢量做相关处理得到该声源位置的判决函数值，并遍历所有可能的声源位置，得到关于全部声源位置的判决函数如下：Step 103-1) Correlate the integrated received data vector and the integrated expected steering vector to obtain the decision function value of the sound source position, and traverse all possible sound source positions to obtain the decision function for all sound source positions as follows :

P_coh-conv(θ，r，z)=w_e ^H(θ，r，z)Rw_e(θ，r，z)P _coh-conv (θ, r, z) = w _e ^H (θ, r, z)Rw _e (θ, r, z)

步骤103-2声源位置的判决函数P_coh-conv(θ，r，z)的最大值出现的位置即为声源位置的估计值。Step 103-2 The position where the maximum value of the decision function P _coh-conv (θ, r, z) of the sound source position appears is the estimated value of the sound source position.

基于上述方法本发明还提供了一种宽带相干模基信号处理系统，所述系统包含：Based on the above method, the present invention also provides a broadband coherent mode-based signal processing system, the system comprising:

综合接收数据向量获取模块，用于获得B个频点的阵列接收数据，对全频带的数据做波束归一化，相干形成综合接收数据向量；The comprehensive receiving data vector acquisition module is used to obtain the array receiving data of B frequency points, perform beam normalization on the data of the whole frequency band, and coherently form a comprehensive receiving data vector;

综合期望导向矢量获取模块，用于获得B个频点的建模向量，并对全频带的建模向量相干形成综合期望导向矢量；The integrated expectation steering vector acquisition module is used to obtain the modeling vectors of B frequency points, and coherently forms the integrated expectation steering vectors to the modeling vectors of the full frequency band;

综合处理模块，用于依据所述的综合接收数据向量和综合期望导向矢量，得到最优导向矢量；并依据最优导向矢量获得最终的目标定位结果。The comprehensive processing module is used to obtain the optimal steering vector according to the integrated received data vector and the integrated expected steering vector; and obtain the final target positioning result according to the optimal steering vector.

上述技术方案中，所述综合接收数据向量获取模块对接收的B个频点的阵列接收数据进行快速傅里叶变换，获得各个单频点的数据；对B个频点的数据依次进行波束形成和归一化处理，形成综合接收数据向量；累积L个快拍阵列接收数据构造阵列协方差矩阵。In the above technical solution, the comprehensive receiving data vector acquisition module performs fast Fourier transform on the received array receiving data of B frequency points to obtain the data of each single frequency point; perform beamforming on the data of B frequency points in sequence and normalization processing to form a comprehensive received data vector; accumulate L snapshot array received data to construct an array covariance matrix.

上述技术方案中，所述综合接收数据向量获取模块进一步包含：In the above technical solution, the integrated receiving data vector acquisition module further includes:

接收子模块，用于用线阵接收空间信号，得到一个快拍时刻N个阵元的时域信号；The receiving sub-module is used to receive space signals with a line array to obtain time-domain signals of N array elements at a snapshot time;

第一处理子模块，对时域数据做快速傅里叶变换，得到B个频点的阵列数据(X₁,X₂,...,X_B)，其中第i个频点的阵列数据Xi表示如下The first processing sub-module performs fast Fourier transform on the time-domain data to obtain array data (X ₁ , X ₂ ,...,X _B ) of B frequency points, wherein the array data Xi of the i-th frequency point expressed as follows

X_i=[x_i1,x_i2,...,x_iN]^T X _i =[x _i1 ,x _i2 ,...,x _iN ] ^T

第二处理子模块，用于对B个频点的数据做波束形成，得到第i个频点的波束数据X_i，beam为The second processing sub-module is used to perform beamforming on the data of B frequency points, and obtain the beam data X _{i of the i-th frequency point, and the beam} is

X_i，beam=X_i*e_i X _{i, beam} = X _i * e _i

将此波束数据作为每个频点的参考信号(X_1ref,X_2ref，...,X_Bref)，即Take this beam data as the reference signal of each frequency point (X _1ref , X _2ref ,...,X _Bref ), namely

X_iref=X_i，beam X _iref = X _{i, beam}

as follows:

累接子模块，用于将B个频点归一化后的数据

累接起来，形成一个长度为B*N的综合接收数据向量

若记：Accumulation sub-module, used to normalize the data of B frequency points

Tiered together to form a comprehensive received data vector of length B*N

If remember:

其中

第三处理子模块，用于接收L个快拍时刻的数据，分别按步骤101-2）-101-4）得到每个快拍的综合接收数据向量，全部L个快拍的综合接收数据向量记为X_e，all；The third processing sub-module is used to receive the data at the time of L snapshots, and obtain the comprehensive received data vector of each snapshot according to steps 101-2)-101-4) respectively, and the comprehensive received data vectors of all L snapshots denoted as X _{e, all} ;

X_e，all=[X_e1,X_e2,...X_eL]X _{e, all} =[X _e1 ,X _e2 ,...X _eL ]

阵列协方差矩阵形成子模块，用于使用L个快拍的综合接收数据向量形成阵列协方差矩阵The array covariance matrix forming submodule is used to form an array covariance matrix using the integrated received data vectors of L snapshots

$R R = = E E. [[{X x}_{e e,, all all} {X x}_{e e,, all all}^{H h}]] - - - - - - ((22))$

上述技术方案中，所述综合期望导向矢量获取模块遍历声源可能出现的位置，并结合海洋环境参数，输入建模工具得到B个频点的加权向量，将B个加权向量累接起来，得到综合期望导向矢量。In the above technical solution, the comprehensive expected steering vector acquisition module traverses the positions where the sound source may appear, and in combination with the marine environment parameters, inputs the weighted vectors of B frequency points into the modeling tool, and accumulates the B weighted vectors to obtain Composite Expected Steering Vector.

上述技术方案中，所述综合处理模块进一步包含：In the above technical solution, the comprehensive processing module further includes:

相关处理子模块，用于将所述的综合接收数据向量和综合期望导向矢量做相关处理得到并遍历所有可能的声源位置，得到关于声源位置的判决函数如下：The correlation processing sub-module is used to correlate the integrated received data vector and the integrated expected steering vector to obtain and traverse all possible sound source positions, and obtain a decision function about the sound source position as follows:

声源定位模块，用于将声源位置的判决函数P_coh-conv(θ，r，z)的最大值出现的位置确定为声源位置的估计值。The sound source localization module is configured to determine the position where the maximum value of the decision function P _coh-conv (θ, r, z) of the sound source position appears as the estimated value of the sound source position.

本发明的优点在于，在做单频点归一化之的时候，选取波束形成后的数据作为参考，这样就最大限度地提高了参考数据的信噪比。总之，本发明提出了为克服现有宽带信息的非相干处理带来的处理损失，在低信噪比的情况下目标检测能力下降，本发明提出一种波束归一化宽带相干模基处理算法。本发明所述的波束归一化宽带模基处理算法，将每个频点的阵列信号用波束形成后的参考向量归一化，从而降低处理器的最小可检测信噪比，提高检测性能。The advantage of the present invention is that, when normalizing a single frequency point, the beam-formed data is selected as a reference, thus maximizing the signal-to-noise ratio of the reference data. In a word, the present invention proposes to overcome the processing loss brought by the non-coherent processing of the existing broadband information, and the target detection ability decreases under the condition of low signal-to-noise ratio. The present invention proposes a beam-normalized broadband coherent mode-based processing algorithm . The beam-normalized broadband modulus processing algorithm of the present invention normalizes the array signal of each frequency point with the reference vector after beamforming, thereby reducing the minimum detectable signal-to-noise ratio of the processor and improving the detection performance.

附图说明 Description of drawings

图1是未采用本发明时，模基定位核心信号处理方法的示意图；Fig. 1 is when not adopting the present invention, the schematic diagram of the core signal processing method of model-based positioning;

图2是采用本发明后，模基定位核心信号处理方法的示意图；Fig. 2 is after adopting the present invention, the schematic diagram of the core signal processing method of module base positioning;

图3是本发明的算法详细处理流程图；Fig. 3 is the detailed processing flowchart of algorithm of the present invention;

图4是采用本发明、未采用本发明时，实施例的常规模基定位方法的输出对比；Fig. 4 is the output comparison of the conventional scale positioning method of the embodiment when the present invention is adopted or not adopted;

图5是采用本发明、未采用本发明时，实施例的自适应模基定位方法的输出对比；Fig. 5 is the output comparison of the adaptive model base positioning method of the embodiment when the present invention is adopted and the present invention is not adopted;

图6是实施例中未采用本发明时，非相干模基定位方法的输出效果图。Fig. 6 is an output effect diagram of the non-coherent mode-based positioning method when the present invention is not used in the embodiment.

图7是实施例中未采用本发明时，相干模基定位方法的输出效果图。Fig. 7 is an output effect diagram of the coherent mode base positioning method when the present invention is not used in the embodiment.

图8是实施例中采用本发明时，波束归一化模基定位方法的输出效果图。Fig. 8 is an output effect diagram of the beam normalized modulo-based positioning method when the present invention is adopted in the embodiment.

具体实施方式 Detailed ways

下面结合附图对本发明进行进一步说明。The present invention will be further described below in conjunction with the accompanying drawings.

一种线列阵声纳，由多个水听器组成，设实际阵元数目为N，共处理B个频点，阵元接收信号表示为x(t)，快拍长度为L。所述的方法包括如下步骤：A linear array sonar is composed of multiple hydrophones. Assuming that the actual number of array elements is N, B frequency points are processed in total. The received signal of the array elements is expressed as x(t), and the snapshot length is L. Described method comprises the steps:

1）用线阵接收空间信号，得到一个快拍时刻N个阵元的时域信号；1) Receive the spatial signal with a line array, and obtain the time-domain signal of N array elements at a snapshot time;

2）对时域数据做快速傅里叶变换，得到B个频点的阵列数据(X₁,X₂,...,X_B)，其中第i个频点的阵列数据Xi表示如下2) Perform fast Fourier transform on the time-domain data to obtain the array data of B frequency points (X ₁ ,X ₂ ,...,X _B ), where the array data Xi of the i-th frequency point is expressed as follows

X_i=[x_i1,x_i2,...,x_iN]^T X _i =[x _i1 ,x _i2 ,...,x _iN ] ^T

3）对B个频点的数据做波束形成，得到第i个频点的波束数据X_i,beam为3) Perform beamforming on the data of B frequency points, and obtain the beam data X _{i, beam} of the i-th frequency point as

X_i,beam=X_i*e_i X _i,beam =X _i *e _i

其中，e_i表示第i个频点的阵列方位补偿向量，由阵元位置、目标方位及信号频率决定。Among them, e _i represents the array orientation compensation vector of the i-th frequency point, which is determined by the array element position, target orientation and signal frequency.

将此波束数据作为每个频点的参考信号(X_1ref,X_2ref,...,X_Bref)，即Take this beam data as the reference signal of each frequency point (X _1ref ,X _2ref ,...,X _Bref ), namely

X_iref=X_i,beam X _iref =X _i,beam

如下式Use this reference signal to normalize the array data (X ₁ ,X ₂ ,...,X _B ) of each frequency point to obtain the normalized data

as follows

4）将B个频点归一化后的数据

累接起来，形成一个长度为B*N的综合接收数据向量

若记4) The data after normalizing the B frequency points

Tiered together to form a comprehensive received data vector of length B*N

Ruoji

其中

表示第j个阵元第i个频点的归一化后的数据，则扩展后的归一化向量in

Represents the normalized data of the i-th frequency point of the j-th array element, then the extended normalized vector

${X x}_{e e 11} = = {[[{\overset{^^}{x x}}_{1111},, {\overset{^^}{x x}}_{1212},, . . . . . .,, {\overset{^^}{x x}}_{11 N N},, {\overset{^^}{x x}}_{21 twenty one},, {\overset{^^}{x x}}_{22 twenty two},, . . . . . .,, {\overset{^^}{x x}}_{22 N N},, . . . . . .,, {\overset{^^}{x x}}_{B B 11},, {\overset{^^}{x x}}_{B B 22},, . . . . . .,, {\overset{^^}{x x}}_{BN BN}]]}^{T T}$

5）接收L个快拍时刻的数据，分别按步骤2-4得到每个快拍的扩展向量，全部L个快拍的综合接收数据向量记为X_e，all；5) Receive the data at the time of L snapshots, and obtain the expansion vector of each snapshot according to steps 2-4 respectively, and denote the comprehensive received data vectors of all L snapshots as X _{e, all} ;

X_e，all=[X_e1,X_e2,...X_eL]X _{e, all} =[X _e1 ,X _e2 ,...X _eL ]

步骤1是接收一个快拍时刻的数据，需要重复L次，步骤5即是这个意思。Step 1 is to receive the data of a snapshot moment, which needs to be repeated L times, and that is what step 5 means.

6）使用L个快拍的综合接收数据向量形成阵列协方差矩阵6) Form an array covariance matrix using the integrated received data vectors of L snapshots

$R R = = E E. [[{X x}_{e e,, all all} {X x}_{e e,, all all}^{H h}]] - - - - - - ((33))$

7）遍历声源可能出现的位置为(θ，r，z)，并结合海洋环境参数，输入建模工具软件包KRAKEN，即可得到B个频点的加权向量(ω₁(θ，r,z)^T,ω₂(θ,r,z)^T,...,ω_B(θ,r,z)^T)，将B个加权向量累接起来，得到长度为B*N的综合期望导向矢量：7) Traverse the possible positions of the sound source as (θ, r, z), and combine the marine environment parameters, input the modeling tool software package KRAKEN, and then get the weighted vector of B frequency points (ω ₁ (θ, r, z) ^T ,ω ₂ (θ,r,z) ^T ,...,ω _B (θ,r,z) ^T ), accumulating B weighted vectors to obtain a comprehensive expectation orientation with length B*N Vector:

w_e(θ，r，z)=[ω₁(θ，r，z)，ω₂(θ，r，z)，...，ω_B(θ，r，z)]^T (4)w _e (θ, r, z) = [ω ₁ (θ, r, z), ω ₂ (θ, r, z), ..., ω _B (θ, r, z)] ^T (4)

该步骤需要为声源可能出现的范围划分网格，设方位、距离、深度对应的网格数为N_θ,N_r,N_z，则需要的建模次数为N_θ*N_r*N_z This step needs to divide the grid for the range where the sound source may appear. Let the number of grids corresponding to the azimuth, distance, and depth be N _θ , N _r , N _z , and the number of modeling required is N _θ *N _r *N _z

8）根据(3)式和(4)式，做相关处理，遍历所有可能的声源位置θ，r,z，即可得到关于声源位置的判决函数8) According to formulas (3) and (4), do correlation processing, traverse all possible sound source positions θ, r, z, and then get the decision function of the sound source position

P_coh-conv(θ，r，z)=w_e ^H(θ，r，z)Rw_e(θ，r，z) (5)P _coh-conv (θ, r, z) = w _e ^H (θ, r, z)Rw _e (θ, r, z) (5)

P_coh-conv(θ，r，z)的最大值出现的位置即为声源位置的估计值。The position where the maximum value of P _coh-conv (θ, r, z) appears is the estimated value of the sound source position.

本发明的基本构思如图2所示：将每个频点的数据累接起来，形成一个综合接收数据的接收阵向量，为了实现频率间的相干处理，对每个频点以一个参考数归一化，本发明中为了尽量提高参考数的信噪比，采用的是波束后的数据。The basic concept of the present invention is shown in Figure 2: the data of each frequency point is accumulated and connected to form a receiving matrix vector of comprehensive received data, in order to realize coherent processing between frequencies, each frequency point is normalized with a reference number First, in the present invention, in order to improve the signal-to-noise ratio of the reference data as much as possible, the data after the beam is used.

所述的方法包括如下步骤，如图3所示：Described method comprises the following steps, as shown in Figure 3:

步骤1：对应图3中的101，用线阵接收空间信号，得到N个阵元的时域信号；第i个水听器接收到的频率ω处的信号可以写为Step 1: Corresponding to 101 in Figure 3, use the line array to receive the space signal, and obtain the time domain signal of N array elements; the signal at the frequency ω received by the i-th hydrophone can be written as

${x x}_{w w}^{i i} = = S S ((w w)) {G G}^{i i} (({r r}_{i i},, {z z}_{i i},, {B B}_{s the s},, {Z Z}_{s the s},, w w)) + + {Q Q}_{w w}^{i i} - - - - - - ((66))$

其中，S(ω)为声源频谱，Gⁱ(r_i，z_i，R_s，Z_s，ω)是声源与第i个水听器间的Green函数，

为该频率处的噪声分量。Among them, S(ω) is the sound source spectrum, G ⁱ (r _i , zi , R _s _, Z _s , ω) is the Green function between the sound source and the i-th hydrophone,

is the noise component at this frequency.

则N元水听器阵列接收到的信号为Then the signal received by the N-element hydrophone array is

${x x}_{w w} = = {[[{x x}_{}^{w w},, {x x}_{}^{w w},, . . . . . .,, {x x}_{w w}^{N N}]]}^{T T} - - - - - - ((77))$

步骤2：对应图1中的102，对时域数据做快速傅里叶变换，得到感兴趣的B个频点的阵列数据(X₁,X₂,...,X_B)；Step 2: Corresponding to 102 in Figure 1, fast Fourier transform is performed on the time domain data to obtain the array data (X ₁ , X ₂ ,..., X _B ) of B frequency points of interest;

$x x ((t t)) &DoubleRightArrow; &DoubleRightArrow; X x ((ω ω)) - - - - - - ((88))$

步骤3：对应图1中的103，对B个频点的数据做波束形成，得到每个频点的参考信号(X_1ref，X_2ref,...,X_Bref)，以此对每个频点的阵列数据做归一化，得到归一化后的数据

例如，若未归一化之前频点w的数据向量为Step 3: Corresponding to 103 in Figure 1, perform beamforming on the data of B frequency points to obtain the reference signal (X _1ref , X _2ref ,...,X _Bref ) of each frequency point, and use this for each frequency point Normalize the array data of points to get the normalized data

For example, if the data vector of frequency point w before normalization is

${x x}_{w w} = = [\begin{matrix} | | S S ((w w)) {G G}_{}^{w w} | | {e e}^{j j} ((arg arg ((S S ((ω ω)))) + + arg arg (({G G}_{w w}^{11})))) \\ . . \\ . . \\ . . \\ | | S S ((w w)) {G G}_{w w}^{N N} | | {e e}^{j j ((arg arg ((S S ((w w)))) + + arg arg (({G G}_{w w}^{N N}))))} \end{matrix}] - - - - - - ((99))$

则归一化后频点w的数据向量可表示为Then the data vector of frequency point w after normalization can be expressed as

步骤4：对应图1中的104，将B个频点归一化后的数据累接起来，形成一个长度为B*N的综合接收数据向量

Step 4: Corresponding to 104 in Figure 1, accumulate and connect the normalized data of B frequency points to form a comprehensive received data vector with a length of B*N

步骤5：对应图1中的105，接收L个快拍的数据，得到每个快拍的综合接收数据向量；Step 5: Corresponding to 105 in Figure 1, receive the data of L snapshots, and obtain the comprehensive received data vector of each snapshot;

步骤6：对应图1中的106，使用L个快拍的综合接收数据向量形成阵列协方差矩阵Step 6: Corresponding to 106 in Figure 1, use the comprehensive received data vectors of L snapshots to form an array covariance matrix

$R R = = E E. [[{X x}_{e e} {X x}_{e e}^{H h}]] - - - - - - ((1111))$

步骤7：对应图1中的107，假设声源的位置为(θ，r，z)，根据海洋环境建模得到B个频点的加权向量，建模得到接收阵列的加权向量，并将B个加权向量累接起来，得到长度为B*N的综合期望导向矢量Step 7: Corresponding to 107 in Figure 1, assuming that the position of the sound source is (θ, r, z), the weighted vectors of B frequency points are obtained according to the marine environment modeling, and the weighted vectors of the receiving array are obtained by modeling, and B Weighted vectors are accumulated and connected to obtain a comprehensive expected steering vector with a length of B*N

w_e=[w₁，w₂，...，w_B]^T (12)w _e =[w ₁ ，w ₂ ，...，w _B ] ^T (12)

${w w}_{i i} = = \frac{{[[G G (({r r}_{11},, {z z}_{11},, {θ θ}_{s the s},, {r r}_{s the s},, {z z}_{s the s})),, G G (({r r}_{22},, {z z}_{22},, {θ θ}_{s the s},, {r r}_{s the s},, {z z}_{s the s})),, . . . . . .,, G G (({r r}_{N N},, {z z}_{N N},, {θ θ}_{s the s},, {r r}_{s the s},, {z z}_{s the s}))]]}^{T T}}{| | [[G G (({r r}_{11},, {z z}_{11},, {θ θ}_{s the s},, {r r}_{s the s},, {z z}_{s the s})),, G G (({r r}_{22},, {z z}_{22},, {θ θ}_{s the s},, {r r}_{s the s},, {z z}_{s the s})),, . . . . . .,, G G (({r r}_{N N},, {z z}_{N N},, {θ θ}_{s the s},, {r r}_{s the s},, {z z}_{s the s}))]] | |} - - - - - - ((1313))$

步骤8：对应图1中的108，根据(11)式和(12)式，使用常规信号处理方法，遍历所有可能的声源位置θ，r,z，即得到关于目标位置的判决函数Step 8: Corresponding to 108 in Figure 1, according to (11) and (12), using conventional signal processing methods, traverse all possible sound source positions θ, r, z, and obtain the decision function about the target position

P_coh-conv(θ，r，z)=w_e ^H(θ，r，z)Rw_e(θ，r，z) (14)P _coh-conv (θ, r, z) = w _e ^H (θ, r, z)Rw _e (θ, r, z) (14)

使用导向矢量自适应优化技术，遍历所有可能的声源位置θ,r,z，即得到自适应信号处理方法时，目标位置的判决函数为：Use the guide vector adaptive optimization technology to traverse all possible sound source positions θ, r, z, that is, when the adaptive signal processing method is obtained, the decision function of the target position is:

${P P}_{coh coh - - mvdr mvdr} ((θ θ,, r r,, z z)) = = \frac{11}{{w w}_{e e}^{H h} ((θ θ,, r r,, z z)) {R R}^{- - 11} {w w}_{e e} ((θ θ,, r r,, z z))} - - - - - - ((1515))$

上述P_coh-conv(θ，r，z)、P_coh-mvdr(θ，r，z)的最大值出现的位置即为声源位置的估计值The position where the maximum value of P _coh-conv (θ, r, z) and P _coh-mvdr (θ, r, z) above appear is the estimated value of the sound source position

此外，上述两种判决函数的不同是由图2（304）步骤决定的，如果（304）步骤不采取优化措施，则得到P_coh-conv(θ，r，z)，如果（304）按照一定的优化准则和约束条件（比如最大阵增益、最优信噪比或者鲁棒自适应技术等）进行优化，则得到P_coh-mvdr(θ，r，z)。In addition, the difference between the above two decision functions is determined by the step (304) in Figure 2. If (304) step does not take optimization measures, P _coh-conv (θ, r, z) will be obtained. If (304) follows a certain The optimization criteria and constraints (such as maximum array gain, optimal signal-to-noise ratio or robust adaptive technology, etc.) are optimized to obtain P _coh-mvdr (θ, r, z).

实施例Example

下面结合数据仿真和附图对本发明的具体实施方式做进一步的详细描述。The specific implementation manner of the present invention will be further described in detail below in combination with data simulation and accompanying drawings.

仿真条件：10元基阵位于海底，阵元间距5m，取典型浅海波导环境，海深30m，等声速梯度，声源频带400-600Hz，取其中的5个频点。得到分别使用常规处理方法和MVDR处理方法时，宽带非相干处理方法（inc）、宽带相干处理方法（MP）、改进的宽带相干处理方法（MP-beam）峰值/背景比值随不同输入信噪比的变化曲线，如图4-图5所示。Simulation conditions: The 10-element array is located on the seabed, the array element spacing is 5m, a typical shallow sea waveguide environment is taken, the sea depth is 30m, the sound velocity gradient is equal, the sound source frequency band is 400-600Hz, and 5 frequency points are taken. When the conventional processing method and the MVDR processing method are used respectively, the peak/background ratio of the broadband non-coherent processing method (inc), the broadband coherent processing method (MP), and the improved broadband coherent processing method (MP-beam) varies with different input SNR The change curve of , as shown in Figure 4-Figure 5.

可以看到，随折信噪比的降低，采用本发明的优势越明显。使用常规信号处理方法（即不做304的优化）的处理结果，如图4所示，高信噪比时，本发明的性能介于两种传统方法之间，低信噪比时，本发明的性能优于两种传统方法。自适应信号处理方法的情况，如图5所示，在各种信噪比情况下，本发明的性能始终优于两种传统方法。It can be seen that the advantages of the present invention are more obvious as the signal-to-noise ratio decreases. Using the processing result of the conventional signal processing method (i.e. without the optimization of 304), as shown in Figure 4, when the signal-to-noise ratio is high, the performance of the present invention is between the two traditional methods, and when the signal-to-noise ratio is low, the present invention outperforms the two traditional methods. In the case of the adaptive signal processing method, as shown in Fig. 5, the performance of the present invention is always better than the two traditional methods in various SNR situations.

图6-图8给出了输入信噪比-5dB情况下，三种自适应模基定位方法的输出结果，可以看到，本发明的处理结果（图8）明显优于两种传统方法。Figures 6-8 show the output results of the three adaptive model-based positioning methods when the input signal-to-noise ratio is -5dB. It can be seen that the processing results of the present invention (Figure 8) are significantly better than the two traditional methods.

总之，本发明相比于传统的非相干处理的方法以及阵元归一化相干处理的方法，具有更优的检测性能。In a word, compared with the traditional non-coherent processing method and the array element normalized coherent processing method, the present invention has better detection performance.

最后所应说明的是，以上实施例仅用以说明本发明的技术方案而非限制。尽管参照实施例对本发明进行了详细说明，本领域的普通技术人员应当理解，对本发明的技术方案进行修改或者等同替换，都不脱离本发明技术方案的精神和范围，其均应涵盖在本发明的权利要求范围当中。Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than limit them. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent replacements to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all of them should be included in the scope of the present invention. within the scope of the claims.

Claims

1. A broadband coherent mode-based signal processing method, the method comprising:

Step 101) Obtain the array reception data of B frequency points, perform beam normalization on the data of the whole frequency band, and coherently form a comprehensive reception data vector, that is, a reception data vector containing information of the whole frequency band;

Step 102) Obtain the modeling vectors of B frequency points, and coherently form the integrated expected steering vector with the modeling vectors of the full frequency band, that is, the expected steering vector containing the information of the full frequency band;

Step 103) Obtain the optimal steering vector according to the integrated received data vector and the integrated expected steering vector;

Step 104) Obtain the final target positioning result according to the optimal steering vector.

2. The broadband coherent mode-based signal processing method according to claim 1, wherein the step 101) is specifically: perform fast Fourier transform on the received array received data of B frequency points to obtain each single The data of frequency points; the data of B frequency points are sequentially processed by beamforming and normalization to form a comprehensive received data vector; the array received data of L snapshots is accumulated to construct an array covariance matrix.

3. The broadband coherent mode-based signal processing method according to claim 1, wherein the step 101) further comprises the following sub-steps:

Step 101-1) Receive spatial signals with a line array to obtain time domain signals of N array elements at a snapshot time;

Step 101-2) Perform fast Fourier transform on the time-domain data to obtain the array data of B frequency points (X ₁ , X ₂ ,...,X _B ), where the array data Xi of the i-th frequency point represents as follows

X _i =[x _i1 ,x _i2 ,...,x _iN ] ^T

Among them, x _ij represents the received data of the i-th frequency point of the j-th array element, and T represents the transpose;

Step 101-3) Perform beamforming on the data of B frequency points, and obtain the beam data Xi _{, beam} of the i-th frequency point as

X _i,beam =C _i *e _i

Among them, e _i represents the array orientation compensation vector of the i-th frequency point, which is determined by the array element position, target orientation and signal frequency;

Take this beam data as the reference signal of each frequency point (X _1ref , X _2ref ,...,X _Bref ), namely

X _iref =X _i,beam

Use this reference signal to normalize the array data (X ₁ ,X ₂ ,...,X _B ) of each frequency point to obtain the normalized data

as follows:

(({\overset{^^}{X x}}_{11},, {\overset{^^}{X x}}_{22},, . . . . . .,, {\overset{^^}{X x}}_{B B})) = = ((\frac{{X x}_{11}}{{X x}_{11 ref ref}},, \frac{{X x}_{22}}{{X x}_{22 ref ref}},, . . . . . .,, \frac{{X x}_{B B}}{{X x}_{Bref Bref}}))

Step 101-4) Normalized data of B frequency points

Tiered together to form a comprehensive received data vector of length B*N

If remember:

{\overset{^^}{X x}}_{11} = = {[[{\overset{^^}{x x}}_{1111},, {\overset{^^}{x x}}_{1212},, . . . . . .,, {\overset{^^}{x x}}_{11 N N}]]}^{T T}

in

{X x}_{e e 11} = = {[[{\overset{^^}{x x}}_{1111},, {\overset{^^}{x x}}_{1212},, . . . . . .,, {\overset{^^}{x x}}_{11 N N},, {\overset{^^}{x x}}_{21 twenty one},, {\overset{^^}{x x}}_{22 twenty two},, . . . . . .,, {\overset{^^}{x x}}_{22 N N},, . . . . . .,, {\overset{^^}{x x}}_{B B 11},, {\overset{^^}{x x}}_{B B 22},, . . . . . .,, {\overset{^^}{x x}}_{BN BN}]]}^{T T};;

Step 101-5) Receive the data of L snapshot moments, respectively follow steps 101-2)-101-4) to obtain the comprehensive received data vector of each snapshot, and record the comprehensive received data vector of all L snapshots as X _{e, all} ;

X _{e, all} =[X _e1 ,X _e2 ,...X _eL ]

The step 101-1) is to receive the data of a snapshot moment, which needs to be repeated L times, and the step 101-5) means this;

Step 101-6) Use the integrated received data vectors of L snapshots to form an array covariance matrix

R R = = E E. [[{X x}_{e e,, all all} {X x}_{e e,, all all}^{H h}]] - - - - - - ((11))

Among them, E means statistical average, and H means conjugate transpose.

4. The broadband coherent mode-based signal processing method according to claim 1, characterized in that, the step 102) specific steps are: traversing the positions where the sound source may appear, and combining the marine environment parameters, inputting the modeling tool to obtain B The weighted vectors of B frequency points are accumulated and connected together to obtain the comprehensive expectation steering vector.

5. The broadband coherent mode-based signal processing method according to claim 1, wherein the step 103) further comprises:

Step 103-1) Correlate the integrated received data vector and the integrated expected steering vector to obtain the decision function value of the sound source position, and traverse all possible sound source positions to obtain the decision function for all sound source positions as follows :

P _coh-conv (θ, r, z) = w _e ^H (θ, r, z)Rw _e (θ, r, z)

Step 103-2 The position where the maximum value of the decision function P _coh-conv (θ, r, z) of the sound source position appears is the estimated value of the sound source position.

6. A broadband coherent mode-based signal processing system, said system comprising:

The comprehensive receiving data vector acquisition module is used to obtain the array receiving data of B frequency points, perform beam normalization on the data of the whole frequency band, and coherently form a comprehensive receiving data vector;

The integrated expectation steering vector acquisition module is used to obtain the modeling vectors of B frequency points, and coherently forms the integrated expectation steering vectors to the modeling vectors of the full frequency band;

The comprehensive processing module is used to obtain the optimal steering vector according to the integrated received data vector and the integrated expected steering vector; and obtain the final target positioning result according to the optimal steering vector.

7. broadband coherent mode-based signal processing system according to claim 6, is characterized in that, described comprehensive reception data vector acquisition module carries out fast Fourier transform to the array reception data of the B frequency points of reception, obtains each single The data of frequency points; the data of B frequency points are sequentially subjected to beamforming and normalization processing to form a comprehensive received data vector; the received data of L snapshot arrays are accumulated to construct an array covariance matrix.

8. The wideband coherent mode-based signal processing system according to claim 6, wherein the integrated receiving data vector acquisition module further comprises:

The receiving sub-module is used to receive space signals with a line array to obtain time-domain signals of N array elements at a snapshot time;

The first processing sub-module performs fast Fourier transform on the time-domain data to obtain array data (X ₁ , X ₂ ,...,X _B ) of B frequency points, wherein the array data Xi of the i-th frequency point expressed as follows

X _i =[x _i1 ,x _i2 ,...,x _iN ] ^T

The second processing sub-module is used to perform beamforming on the data of B frequency points, and obtain the beam data Xi _{, beam} of the i-th frequency point as

X _{i, beam} = X _i * e _i

Take this beam data as the reference signal of each frequency point (X _1ref ,X _2ref ,...,X _Bref ), namely

X _iref =X _i,beam

Use this reference signal to normalize the array data (X ₁ ,X ₂ ,...,X _B ) of each frequency point to obtain the normalized data as follows:

(({\overset{^^}{X x}}_{11},, {\overset{^^}{X x}}_{22},, . . . . . .,, {\overset{^^}{X x}}_{B B})) = = ((\frac{{X x}_{11}}{{X x}_{11 ref ref}},, \frac{{X x}_{22}}{{X x}_{22 ref ref}},, . . . . . .,, \frac{{X x}_{B B}}{{X x}_{Bref Bref}}))

Accumulation sub-module, used to normalize the data of B frequency points Tiered together to form a comprehensive received data vector of length B*N

If remember:

{\overset{^^}{X x}}_{11} = = {[[{\overset{^^}{x x}}_{1111},, {\overset{^^}{x x}}_{1212},, . . . . . .,, {\overset{^^}{x x}}_{11 N N}]]}^{T T}

in

{X x}_{e e 11} = = {[[{\overset{^^}{x x}}_{1111},, {\overset{^^}{x x}}_{1212},, . . . . . .,, {\overset{^^}{x x}}_{11 N N},, {\overset{^^}{x x}}_{21 twenty one},, {\overset{^^}{x x}}_{22 twenty two},, . . . . . .,, {\overset{^^}{x x}}_{22 N N},, . . . . . .,, {\overset{^^}{x x}}_{B B 11},, {\overset{^^}{x x}}_{B B 22},, . . . . . .,, {\overset{^^}{x x}}_{BN BN}]]}^{T T};;

The third processing sub-module is used to receive the data at the time of L snapshots, and obtain the comprehensive received data vector of each snapshot according to steps 101-2)-101-4) respectively, and the comprehensive received data vectors of all L snapshots denoted as X _{e, all} ;

X _{e, all} =[X _e1 ,X _e2 ,...X _eL ]

The array covariance matrix forming submodule is used to form an array covariance matrix using the integrated received data vectors of L snapshots

R R = = E E. [[{X x}_{e e,, all all} {X x}_{e e,, all all}^{H h}]] - - - - - - ((22))

Among them, E means statistical average, and H means conjugate transpose.

9. broadband coherent mode-based signal processing system according to claim 6, is characterized in that, the position that described integrated expectation steering vector acquisition module traverses possible occurrence of sound source, and in conjunction with marine environment parameter, input modeling tool obtains B The weighted vectors of the frequency points are accumulated and connected to the B weighted vectors to obtain the integrated expected steering vector.

10. The broadband coherent mode-based signal processing system according to claim 6, wherein the integrated processing module further comprises:

The correlation processing sub-module is used to correlate the integrated received data vector and the integrated expected steering vector to obtain and traverse all possible sound source positions, and obtain a decision function about the sound source position as follows:

P _coh-conv (θ, r, z) = w _e ^H (θ, r, z)Rw _e (θ, r, z)

The sound source localization module is configured to determine the position where the maximum value of the decision function P _coh-conv (θ, r, z) of the sound source position appears as the estimated value of the sound source position.