CN-121978666-A - Knowledge priori and auxiliary sample based towed sonar space-time reverberation suppression method

CN121978666ACN 121978666 ACN121978666 ACN 121978666ACN-121978666-A

Abstract

The invention discloses a towed sonar space-time reverberation suppression method based on knowledge prior and auxiliary samples, which comprises the steps of acquiring the auxiliary samples and samples to be detected based on a sonar array, uniformly dividing a space-time plane along a space frequency axis and a Doppler axis, constructing a space-time guide vector set, screening vectors positioned in a reverberation energy concentration area according to the space-time guide vector set to obtain a first set, carrying out energy analysis and effective vector screening according to the auxiliary samples and the first set to obtain a second set, distributing sparse punishment weights to each vector in the second set, constructing an optimization model containing data fitting items and weighted sparse constraint items according to the second set, solving the optimization model to obtain a reverberation coefficient vector, obtaining a reconstructed reverberation component according to the reverberation sparse coefficient vector and the second set, subtracting the reconstructed reverberation component from the samples to be detected, and finishing reverberation suppression.

Inventors

HONG LERONG
ZHU YALONG
Fan Zetong
ZHOU XIAOPING
WANG LEI
LIAO SHUHAN
LUO AN

Assignees

湖南大学

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (8)

1. A towed sonar space-time reverberation suppression method based on knowledge prior and auxiliary samples is characterized by comprising the following steps: The method comprises the steps of acquiring an auxiliary sample and a sample to be detected based on a sonar array, uniformly dividing a space-time plane along a space frequency axis and a Doppler axis, constructing a space-time guide vector set, screening vectors positioned in a reverberation energy concentration area according to the space-time guide vector set to obtain a first set, and carrying out energy analysis and effective vector screening according to the auxiliary sample and the first set to obtain a second set; An optimization model containing a data fitting term and a weighted sparse constraint term is built according to the second set, a reverberation sparse coefficient vector is obtained by solving the optimization model, a reconstructed reverberation component is obtained according to the reverberation sparse coefficient vector and the second set, and the reconstructed reverberation component is subtracted from the sample to be detected to complete reverberation suppression.
2. The method for drag sonar space-time reverberation suppression based on knowledge prior and auxiliary samples according to claim 1, wherein the step of constructing the space-time steering vector set after uniformly dividing the space-time plane along the space frequency axis and the doppler axis comprises the steps of: And uniformly dividing a space-time plane along a space frequency axis and a Doppler axis according to the motion parameters and the array parameters of the sonar platform, and constructing a space-time guide vector for each grid point to obtain a space-time guide vector set, wherein the space-time guide vector consists of a Kronecker product of a space guide component and a time guide component.
3. The method for drag sonar space-time reverberation suppression based on knowledge prior and auxiliary samples according to claim 2, wherein the screening vectors located in a reverberation energy concentration region according to the space-time steering vector set to obtain a first set includes: Based on the navigation speed and yaw angle of the sonar platform, the frequency of the emission signal and the sound velocity, a geometric relation model between the reverberation center frequency and the angle is established, and for a given angle The geometric relationship model is expressed as: ; Wherein, the Is the ship speed; Is a yaw angle; for the transmit signal frequency; is the sound velocity; And Respectively representing a port reverberation center frequency and a starboard reverberation center frequency; defining a reverberant energy concentration area in an angle-Doppler plane according to the geometric relationship model Comprising: ; Wherein, the Is the spatial frequency axis The discrete angles to which the individual grids correspond, , Space frequency axis grid number of space time plane; is Doppler axis The doppler frequencies corresponding to the individual grids, , The Doppler axis grid number is the space-time plane; the reverberation stretching range is preset; Screening the region positioned in the reverberation energy concentration region according to the space-time guide vector set Is used to obtain a first set.
4. The method for drag sonar space-time reverberation suppression based on knowledge prior and auxiliary samples according to claim 3, wherein performing energy analysis and effective vector screening according to the auxiliary samples and the first set to obtain a second set comprises: Analyzing the matching energy of each vector in the first set according to the auxiliary sample to obtain an energy statistical result: ; Wherein, the Representing a first set Middle (f) The energy statistics of the individual vectors; representing a first set The first of (3) A vector; represent the first Frame-assisted samples; Representing the total number of auxiliary samples; represents a conjugate transpose; Screening vectors with the energy statistical result being greater than or equal to a reverberation energy threshold value in the first set to obtain a second set: ; Wherein, the Is a reverberation energy threshold; Is the second set.
5. The method for drag sonar space-time reverberation suppression based on knowledge prior and auxiliary samples according to claim 4, wherein assigning sparse penalty weights to each vector in the second set comprises: Sparse penalty weights assigned to each vector in the second set Expressed as: ; Wherein, the Is a smoothing parameter; parameters are adjusted for the weights.
6. The method for drag sonar space-time reverberation suppression based on knowledge a priori and auxiliary samples according to claim 5, wherein constructing an optimization model including data fitting terms and weighted sparse constraint terms from the second set includes: according to the second set and the sparse penalty weight, an optimization model comprising a data fitting term and a weighted sparse constraint term is constructed for the sample to be tested, and the optimization model is expressed as follows: ; Wherein, the Representing a matrix corresponding to the second set as a second matrix; representing a sample to be tested; Representing a reverberation sparse coefficient vector; Is a regularization parameter; represent the first Values of the individual reverberation sparse coefficient vector elements.
7. The method for drag sonar space-time reverberation suppression based on knowledge prior and auxiliary samples according to claim 6, wherein solving the optimization model to obtain a reverberation sparse coefficient vector comprises: solving the optimization model by adopting an alternate direction multiplier method, comprising: Introducing auxiliary variables Converting the optimization model into an optimization problem with equality constraints: ; Wherein, the , Representation of The first of (3) A component; Representing the square of the vector L2 norm; The auxiliary variable Updating based on the soft threshold contraction operator, expressed as: ; Wherein, the Is the first Updated by a second iteration Is a value of (2); Is the first Updated by a second iteration Is used as a reference to the value of (a), In order to pair the vector of variables, , Is that Corresponds to the first Components of the individual reverberation sparse coefficient constraint terms; Is the first Updated by a second iteration Is a value of (2); For the soft threshold puncturing operator, ; Is a punishment parameter; Punishment of weight vectors for sparsity; For the first Auxiliary variable in secondary iteration update Is the first of (2) Individual components The calculation method comprises the following steps: ; ; Wherein, the Is that A corresponding shrink threshold; an extremely small positive number to prevent division by zero; Is that Is the first of (2) A component; Is that Is the first of (2) A component; For the following The updating method comprises the following steps: ; Wherein, the Is the first Updated by a second iteration Is a value of (2); For the dual variable vector The consistency adjustment is carried out, and the updating method comprises the following steps: ; Wherein, the Is the first Updated by a second iteration Is a value of (2); For a pair of 、 And Performing cyclic and alternative updating until the original variable And auxiliary variables The error between the two satisfy the convergence condition to obtain the reverberation sparse coefficient vector 。
8. The method for drag sonar space-time reverberation suppression based on knowledge prior and auxiliary samples according to claim 7, wherein obtaining a reconstructed reverberation component according to the reverberation sparse coefficient vector and the second set, subtracting the reconstructed reverberation component from the sample to be detected, and completing reverberation suppression comprises: The reverberation sparse coefficient vector is set A matrix corresponding to the second set, namely the second matrix Multiplying to obtain a reconstructed reverberation component And then the sample to be tested is processed Subtracting the reconstructed reverberant component Obtaining residual signals with reverberation suppression Expressed as: 。

Description

Knowledge priori and auxiliary sample based towed sonar space-time reverberation suppression method Technical Field The invention relates to the field of underwater sound signal processing, in particular to a towed sonar space-time reverberation suppression method based on knowledge priori and auxiliary samples. Background In an active sonar system, detection and parameter estimation of an underwater target are realized by transmitting acoustic signals and processing echo signals. In shallow sea and complex marine environments, sonar echo signals typically contain a large number of reverberant components caused by non-uniform media on the ocean floor, sea surface, and water. The reverberation has the characteristics of strong energy, long duration, complex statistical characteristics and the like, often submerges weak target echoes, and seriously influences the detection performance and parameter estimation accuracy of a sonar system. Existing sonar reverberation suppression typically uses a STAP (Space-time adaptive Processing) method. The STAP method suppresses reverberation and interference by constructing a space-time filter, and can achieve better performance under ideal conditions. However, in an actual shallow sea environment, reverberation often presents obvious non-stationary characteristics, the number of training samples is limited, and the statistical characteristics and data to be tested are mismatched, so that covariance matrix estimation deviation is easily caused, even rank deficiency problem occurs, and the stability and the robustness of an algorithm are affected. Furthermore, conventional STAP methods typically require global filtering of the entire angle-Doppler domain, underutilizing the structured distribution characteristics exhibited by reverberation in the space-time domain, and target echoes may also be attenuated simultaneously during the filtering process. In recent years, sparse representation and compressed sensing theory are introduced into sonar signal processing, and reverberation and target signals are represented as linear combinations of a small number of dictionary atoms by constructing a space-time steering vector dictionary, so that reverberation suppression is realized. The method relieves the problem of difficult covariance matrix estimation under the condition of small samples to a certain extent. However, existing sparse constraint methods typically impose a uniform sparse penalty on all dictionary atoms, failing to fully exploit prior information contained in the auxiliary training data. When the reverberation energy is obviously concentrated in the space-time domain, the uniform sparse constraint is difficult to effectively distinguish the reverberation atoms from the non-reverberation atoms, so that the reverberation suppression performance is limited. In the actual sonar reverberation suppression process, not only the structural distribution characteristic of the reverberation in the space-time domain needs to be fully utilized, but also different space-time guide vectors need to be subjected to differential constraint by combining with statistical information in auxiliary training data so as to improve the accuracy and robustness of reverberation suppression, but the existing method still has defects in the aspects of reverberation structure utilization and training information introduction, and is difficult to realize stable and effective reverberation suppression under complex environment and limited training sample conditions. Therefore, a new technical solution is needed to solve the technical problem of how to effectively suppress the reverberation interference in the sonar echo signal in the complex reverberation background. Disclosure of Invention The invention provides a towed sonar space-time reverberation suppression method based on knowledge priori and auxiliary samples, which is used for solving the technical problem of how to effectively suppress reverberation interference in a sonar echo signal under a complex reverberation background. In order to achieve the above purpose, the invention provides a towed sonar space-time reverberation suppression method based on knowledge priori and auxiliary samples, comprising the following steps: The method comprises the steps of obtaining an auxiliary sample and a sample to be detected based on a sonar array, uniformly dividing a space-time plane along a space frequency axis and a Doppler axis, and then constructing a space-time guide vector set, screening vectors positioned in a reverberation energy concentration area according to the space-time guide vector set to obtain a first set, and carrying out energy analysis and effective vector screening according to the auxiliary sample and the first set to obtain a second set; The method comprises the steps of constructing an optimization model containing a data fitting item and a weighted sparse constraint item according to a second set, solving the optimization