阅读说明：本技术 适用于部分均匀混响环境的空时自适应检测方法及系统 (Space-time self-adaptive detection method and system suitable for partial uniform reverberation environment ) 是由 郝程鹏 闫林杰 刘明刚 侯朝焕 于 2020-11-10 设计创作，主要内容包括：本发明提出了适用于部分均匀混响环境的空时自适应检测方法及系统,所述方法包括：获取空间对称线阵接收的回波数据的待检测数据矩阵和辅助数据矩阵；将待检测数据和辅助数据进行联合,采用最大似然估计对预先构建的检测统计量的参数进行估计；将估计的参数输入预先建立的空时自适应检测器,完成目标的自适应检测。本发明的方法大幅提高了对回波数据的利用率,且与其它同类型的检测方法相比,在小样本辅助数据下具有更好的检测性能和距离估计性能,在实际应用中效率大幅提升。(The invention provides a space-time self-adaptive detection method and a system suitable for a partial uniform reverberation environment, wherein the method comprises the following steps: acquiring a data matrix to be detected and an auxiliary data matrix of echo data received by a space symmetric linear array; combining the data to be detected and the auxiliary data, and estimating parameters of the pre-constructed detection statistics by adopting maximum likelihood estimation; and inputting the estimated parameters into a pre-established space-time adaptive detector to finish the adaptive detection of the target. The method greatly improves the utilization rate of the echo data, has better detection performance and distance estimation performance under small sample auxiliary data compared with other detection methods of the same type, and greatly improves the efficiency in practical application.)

1. A method of space-time adaptive detection applicable to a partially uniform reverberant environment, the method comprising:

acquiring a data matrix to be detected and an auxiliary data matrix of echo data received by a space symmetric linear array;

combining the data to be detected and the auxiliary data, and estimating parameters of the pre-constructed detection statistics by adopting maximum likelihood estimation;

and inputting the estimated parameters into a pre-established space-time adaptive detector to finish the adaptive detection of the target.

2. The space-time adaptive detection method applicable to a partial uniform reverberation environment according to claim 1, wherein the acquiring a to-be-detected data matrix and an auxiliary data matrix of echo data received by a spatially symmetric linear array specifically comprises:

the space symmetrical linear array consists of N array elements, and after receiving echo signals and performing signal processing, a discrete space-time processing N-dimensional echo complex vector z is obtained_iThe received echo vector z to be detected of the ith distance unit_iExpressed as:

z_i＝s_i+n_i∈C^N×1 (1)

wherein C represents a complex field, n_iRepresenting the complex Gaussian interference vector, s_iRepresenting a leaked target signal vector, and when energy leakage exists in a target, signal energy can be leaked into a left adjacent distance unit and a right adjacent distance unit of the target, so that a target energy leakage model consisting of three adjacent distance units is obtained, and s_iExpressed as:

wherein alpha is a complex amplitude factor of a received target echo signal and is an unknown determined parameter; chi shape_p(. is a complex ambiguity function of the transmitted signal; epsilon₀∈[0,T_p]The severity of the target energy leakage, T, can be measured for the remaining time delay_pIs the pulse width; v is a target normalized space-domain steering vector; f is the Doppler shift introduced by the target and v; l represents the serial number of the sample to be detected;

taking the ith distance unit to be detected as the center, delaying the residual time by epsilon ∈ T_p/2,T_p/2]Redefined as:

z_k∈C^N×1representing K uniform auxiliary data collected from neighboring range cells of the data to be detected, containing only white noiseAnd reverberation;

Z_L＝[z_l-1,z_l,z_l+1]∈C^N×3for data matrix to be detected, Z_K＝[z₁,...,z_K]∈C^N×KFor the auxiliary data matrix, Z ═ Z_L,Z_K]∈C^N×(3+K)Is a joint data matrix.

3. A space-time adaptive detection method suitable for a partial uniform reverberation environment according to claim 2, characterized in that the data to be detected and the auxiliary data are combined, and the parameters of the pre-constructed detection statistics are estimated by maximum likelihood estimation; the method specifically comprises the following steps:

step 2-1), constructing a detection statistic T based on a GLRT criterion:

wherein, gamma is more than 0 to represent unknown energy scale factor, and M is a covariance matrix of the symmetric linear array; v is a normalized space-domain steering vector, and M and v have the characteristic of oblique symmetry, namely:

M＝J_NM^*J_N,v＝J_Nv^*

wherein, J_N∈R^N×NIs a permutation matrix (·)^*For conjugate operations, R is the real number domain, where the permutation matrix J_N∈R^N×NThe matrix is a square matrix with an oblique diagonal line of 1 and other elements of 0;

step 2-2) oblique symmetry characteristics based on M and v, f_j(Z; j ε, j α, γ, M) is represented by_jJ is 0,1 assumes the probability density function of the joint data matrix Z:

where det (-) represents the determinant operation of the matrix and tr (-) represents the trace of the matrix^TAnd (·)^HRepresenting the transpose and conjugate transpose of the matrix, the intermediate variable S and the interference data matrix F (j α) are:

x is a data matrix generated by data to be detected:

vectorAndcomprises the following steps:

the fuzzy function matrix D is:

in the formula, t₁,t₂,t₃All are time delays: t is t₁＝-T_p-ε,t₂＝-ε,t₃＝T_p-ε；

Step 2-3) adopting maximum likelihood estimation to H_jM under the assumption that j is 0,1 is estimated to obtain an estimation resultComprises the following steps:

substituting (9) into equation (6), the detection statistic is:

step 2-4) is based on formula (10), and the maximum likelihood estimation of α is:

solving the first-order partial derivative of alpha and setting zero for formula (11), and obtaining the estimation result of alpha as:

wherein Q is an intermediate variable matrix:denotes S^-1/2v's spanned subspace matrixOrthogonal complement of (I)_NIs an N-dimensional unit matrix;

step 2-5) willSubstitution (10), solve at H₁Maximum likelihood estimation assuming gammaComprises the following steps:

the intermediate matrices B and C are:

where the matrix Q can be characterized as Q ═ U (γ)₁I₆+Λ)U^HWherein U is equal to C^6×6Is a unitary matrix with lambda being a characteristic value of lambda₁,...,λ₆A diagonal matrix of (c);

the intermediate matrices E and G are:

matrices W and V are:

substituting the characteristic decomposition of Q into (14),

wherein the content of the first and second substances,d＝c^*， m＝p^*，n＝q^*；

to h₁(γ₁) To obtain gamma₁The first partial derivative of (a) is reset to zero to obtain gamma₁Is estimated value ofSolving the nonlinear equation by using a fsolve function;

steps 2-6) are based on equation (10) at H₀Suppose the following maximum likelihood estimate for γ is:

is simplified to obtain

To X^HS^-1Performing characteristic decomposition on X to obtainWherein U is₀∈C^6×6Being a unitary matrix, Λ₀For a characteristic value of λ_0,1,...,λ_0,6A diagonal matrix of (c); after substitution (19), the result is simplified:

to h₀(γ₀) To obtain gamma₀The first partial derivative of (a) is set to zero, and the pair gamma is obtained as well₀Is estimated value ofHere, theNumerical methods are required for solving.

4. A space-time adaptive detection method for partial uniform reverberation environments as set forth in claim 3, wherein the space-time adaptive detector is:

wherein H₁And H₀Respectively representing the assumption of target and the assumption of no target; eta represents the corresponding detection threshold under a certain false alarm probability.

5. A space-time adaptive detection method for partial uniform reverberation environment according to claim 4, wherein the estimated parameters are input to a pre-established space-time adaptive detector, and the adaptive detection of the target specifically comprises:

step 3-1) calculating a detection statistic T:

step 3-2) when the test statistic T is greater than the detection threshold eta, testing H₁If yes, the detection result is targeted, otherwise, H is checked₀And if yes, the detection result is no target.

6. A space-time adaptive detection system adapted for use in a partially uniform reverberant environment, the system comprising: the device comprises a pre-established space-time adaptive detector, a data acquisition module, a parameter estimation module and a detection module;

the data acquisition module is used for acquiring a to-be-detected data matrix and an auxiliary data matrix of echo data received by the spatially symmetric linear arrays;

the parameter estimation module is used for combining the data to be detected and the auxiliary data and estimating the parameters of the pre-constructed detection statistics by adopting maximum likelihood estimation;

and the detection module is used for inputting the estimated parameters into the space-time self-adaptive detector to finish the self-adaptive detection of the target.

Technical Field

The invention relates to the technical field of underwater sound, in particular to a space-time self-adaptive detection method and system suitable for a partial uniform reverberation environment.

Background

The underwater target detection technology is used for detecting whether a target exists in an underwater sound field based on a signal detection and estimation theory. In shallow sea active sonar signal detection, reverberation as a major interference factor will directly threaten the target detection performance of the system. Particularly, for sonar carriers with certain motion speed, reverberation in different directions has different doppler frequency shifts, so that a reverberation spectrum is expanded in a doppler-angle domain, and the reverberation spectrum is difficult to be effectively inhibited by traditional methods such as beam forming and matched filtering. In the 90 s of the 20 th century, Jaffer first applied space-time adaptive processing (STAP) techniques based on space-time domain joint filtering to active sonar reverberation suppression. On the basis of the STAP, a space-time adaptive detection (STAD) technology aiming at reverberation suppression and target detection is developed, wherein more classical detection methods mainly comprise generalized likelihood ratio detection (GLRT), adaptive matched filter detection (AMF), Wald detection methods and the like.

However, the conventional STAD method still has two disadvantages: firstly, the estimation and inversion of a high-dimensional interference covariance matrix are needed, and the demand for uniform auxiliary data is large; secondly, an ideal target sampling model is mostly adopted, namely, the sampling point is just consistent with the peak position of the target matched filtering output, and the target energy leakage condition is ignored. In view of the first deficiency, Nitzberg indicates that the objective of reducing the amount of auxiliary data can be achieved by utilizing the oblique symmetry characteristic of the reverberation covariance matrix in the spatially symmetric linear array system. Aiming at the second defect, a detection method based on a target energy leakage sampling model is researched by a Hospital institute of acoustics, a Hospital institute of Helcheng and an Italy scholarand. In recent years, an oblique symmetry AMF (PM-AMF-PHE) detection method based on an oblique symmetry characteristic and a target energy leakage sampling model under a partial uniform reverberation environment has been proposed. The part of the uniform data is that the reverberation covariance matrix structure of the data to be detected and the reverberation covariance matrix structure of the auxiliary data are the same and only differ by one unknown energy scale factor, and researches prove that the environment is closer to the actual working scene of the sonar. The PM-AMF-PHE method effectively compensates the target energy leakage loss and improves the target detection performance under the condition that the quantity of auxiliary data is limited.

In an actual environment, under the influence of factors such as fluctuation of an underwater interface, channel variation and the like, uniform auxiliary data are extremely limited or even cannot be acquired, and therefore a detection method in a partial uniform reverberation environment is provided. Although the existing PM-AMF-PHE method considers the oblique symmetry characteristic of a reverberation covariance matrix and a target energy leakage sampling model when receiving data for modeling, the method adopts a two-step GLRT criterion for designing the detection method, namely, to-be-detected data and auxiliary data are separately used to realize derivation of detection statistics and Maximum Likelihood Estimation (MLE) of unknown parameters, joint utilization of the to-be-detected data and the auxiliary data cannot be realized, the data utilization rate of received echoes is low, and the detection performance is restricted.

The existing PM-AMF-PHE adopts a two-step GLRT criterion, the joint use of data to be detected and auxiliary data cannot be realized in the design process of a detection method, the utilization rate of data of received back waves is low, and the detection performance cannot meet the requirement under the condition that the quantity of the auxiliary data is limited.

Disclosure of Invention

The invention aims to overcome the technical defects and provides a high-performance space-time adaptive detection method suitable for a partial uniform reverberation environment, wherein a target energy leakage sampling model is adopted to make up for leakage loss, the requirement on auxiliary data is reduced by utilizing the oblique symmetry characteristic of a reverberation covariance matrix in a space symmetric linear array, and the unknown parameter estimation and detection statistic derivation are carried out by jointly using data to be detected and the auxiliary data, so that the utilization rate of the received echo data is improved, and the method has obvious target detection advantages and is beneficial to practical application under the condition that the quantity of the auxiliary data is limited.

To achieve the above object, embodiment 1 of the present invention proposes a space-time adaptive detection method suitable for a partial uniform reverberation environment, the method including:

acquiring a data matrix to be detected and an auxiliary data matrix of echo data received by a space symmetric linear array;

combining the data to be detected and the auxiliary data, and estimating parameters of the pre-constructed detection statistics by adopting maximum likelihood estimation;

and inputting the estimated parameters into a pre-established space-time adaptive detector to finish the adaptive detection of the target.

As an improvement of the above method, the acquiring a to-be-detected data matrix and an auxiliary data matrix of echo data received by the spatially symmetric linear array specifically includes:

the space symmetrical linear array consists of N array elements, and after receiving echo signals and performing signal processing, a discrete space-time processing N-dimensional echo complex vector z is obtained_iThe received echo vector z to be detected of the ith distance unit_iExpressed as:

z_i＝s_i+n_i∈C^N×1 (1)

wherein C represents a complex field, n_iRepresenting the complex Gaussian interference vector, s_iRepresenting a leaked target signal vector, and when energy leakage exists in a target, signal energy can be leaked into a left adjacent distance unit and a right adjacent distance unit of the target, so that a target energy leakage model consisting of three adjacent distance units is obtained, and s_iExpressed as:

wherein alpha is a complex amplitude factor of a received target echo signal and is an unknown determined parameter; chi shape_p(. is a complex ambiguity function of the transmitted signal; epsilon₀∈[0,T_p]The severity of the target energy leakage, T, can be measured for the remaining time delay_pIs the pulse width; v is a target normalized space-domain steering vector; f is the Doppler shift introduced by the target and v; l represents the serial number of the sample to be detected;

taking the ith distance unit to be detected as the center, delaying the residual time by epsilon ∈ T_p/2,T_p/2]Redefined as:

z_k∈C^N×1k pieces of uniform auxiliary data collected from adjacent distance units of data to be detected are represented, and only two interference components of white noise and reverberation are contained;

Z_L＝[z_l-1,z_l,z_l+1]∈C^N×3for data matrix to be detected, Z_K＝[z₁,...,z_K]∈C^N×KFor the auxiliary data matrix, Z ═ Z_L,Z_K]∈C^N×(3+K)Is a joint data matrix.

As an improvement of the above method, the data to be detected and the auxiliary data are combined, and the parameters of the pre-constructed detection statistics are estimated by using maximum likelihood estimation; the method specifically comprises the following steps:

step 2-1), constructing a detection statistic T based on a GLRT criterion:

wherein, gamma is more than 0 to represent unknown energy scale factor, and M is a covariance matrix of the symmetric linear array; v is a normalized space-domain steering vector, and M and v have the characteristic of oblique symmetry, namely:

M＝J_NM^*J_N,v＝J_Nv^*

wherein, J_N∈R^N×NIs a permutation matrix (·)^*For conjugate operations, R is the real number domain, where the permutation matrix J_N∈R^N×NThe matrix is a square matrix with an oblique diagonal line of 1 and other elements of 0;

step 2-2) oblique symmetry characteristics based on M and v, f_j(Z; j ε, j α, γ, M) is represented by_jJ is 0,1 assumes the probability density function of the joint data matrix Z:

wherein, det (-) represents the determinant operation of the matrix, tr (-) represents the trace of the matrix, (.)^TAnd (·)^HRepresenting the transpose and conjugate transpose of the matrix, the intermediate variable S and the interference data matrix F (j α) are:

x is a data matrix generated by data to be detected:

vectorAndcomprises the following steps:

the fuzzy function matrix D is:

in the formula, t₁,t₂,t₃All are time delays: t is t₁＝-T_p-ε,t₂＝-ε,t₃＝T_p-ε；

Step 2-3) adopting maximum likelihood estimation to H_jM under the assumption that j is 0,1 is estimated to obtain an estimation resultComprises the following steps:

substituting (9) into equation (6), the detection statistic is:

step 2-4) is based on formula (10), and the maximum likelihood estimation of α is:

solving the first-order partial derivative of alpha and setting zero for formula (11), and obtaining the estimation result of alpha as:

wherein Q is an intermediate variable matrix: denotes S^-1/2v's spanned subspace matrixOrthogonal complement of (I)_NIs an N-dimensional unit matrix;

step 2-5) willSubstitution (10), solve at H₁Maximum likelihood estimation assuming gammaComprises the following steps:

the intermediate matrices B and C are:

where the matrix Q can be characterized as Q ═ U (γ)₁I₆+Λ)U^HWherein U is equal to C^6×6Is a unitary matrix with lambda being a characteristic value of lambda₁,...,λ₆A diagonal matrix of (c);

the intermediate matrices E and G are:

matrices W and V are:

substituting the characteristic decomposition of Q into (14),

wherein the content of the first and second substances, **m＝p^*，n＝q^*；

to h₁(γ₁) To obtain gamma₁The first partial derivative of (a) is reset to zero to obtain gamma₁Is estimated value ofSolving the nonlinear equation by using a fsolve function;

steps 2-6) are based on equation (10) at H₀Suppose the following maximum likelihood estimate for γ is:

is simplified to obtain

To X^HS^-1Performing characteristic decomposition on X to obtainWherein U is₀∈C^6×6Being a unitary matrix, Λ₀For a characteristic value of λ_0,1,...,λ_0,6A diagonal matrix of (c); after substitution (19), the result is simplified:

to h₀(γ₀) To obtain gamma₀The first partial derivative of (a) is set to zero, and the pair gamma is obtained as well₀Is estimated value ofHere, theNumerical methods are required for solving.

As an improvement of the above method, the space-time adaptive detector is:

wherein H₁And H₀Respectively representing the assumption of target and the assumption of no target; eta represents the corresponding detection threshold under a certain false alarm probability.

As an improvement of the above method, the inputting of the estimated parameters into a pre-established space-time adaptive detector to perform adaptive detection of the target specifically includes:

step 3-1) calculating a detection statistic T:

step 3-2) when the test statistic T is greater than the detection threshold eta, testing H₁If yes, the detection result is targeted, otherwise, H is checked₀And if yes, the detection result is no target.

Embodiment 2 of the present invention provides a space-time adaptive detection system suitable for a partial uniform reverberation environment, where the system includes: the device comprises a pre-established space-time adaptive detector, a data acquisition module, a parameter estimation module and a detection module;

the data acquisition module is used for acquiring a to-be-detected data matrix and an auxiliary data matrix of echo data received by the spatially symmetric linear arrays;

the parameter estimation module is used for combining the data to be detected and the auxiliary data and estimating the parameters of the pre-constructed detection statistics by adopting maximum likelihood estimation;

and the detection module is used for inputting the estimated parameters into the space-time self-adaptive detector to finish the self-adaptive detection of the target.

The invention has the advantages that:

the method of the invention jointly uses the data to be detected and the auxiliary data to realize the derivation of the detection statistics and the MLE of all unknown parameters, replaces the traditional derivation method which only uses the data to be detected or the auxiliary data in the two-step GLRT criterion, greatly improves the utilization rate of the echo data, has better detection performance and distance estimation performance under small sample auxiliary data compared with other detection methods of the same type, and greatly improves the efficiency in practical application.

Drawings

FIG. 1 is a flow chart of a space-time adaptive detection method for a partial uniform reverberation environment according to the present invention;

FIG. 2 shows P of different detection methods_dA curve that varies with SRNR;

FIG. 3 shows δ for different detection methods_rmsCurve with SRNR.

Detailed Description

The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

The invention designs an adaptive solution under a partial uniform reverberation environment based on a GLRT (generalized likelihood ratio test) criterion. The sonar system is assumed to adopt a space symmetric linear array to receive echoes, a target signal adopts a target energy leakage sampling model, and a reverberation signal adopts an oblique symmetric prior structure; in consideration of a target energy leakage phenomenon existing in sampling, an energy leakage sampling model is adopted to make up for leakage loss in the process of modeling a received signal, and the oblique symmetry characteristic of a reverberation covariance matrix is utilized to reduce the requirement amount of auxiliary data in the process of modeling a reverberation signal. The derivation process of the detection method only needs one step, namely, the MLE of the parameters is directly realized by combining the data to be detected and the auxiliary data. And finally substituting the estimated value instead of the theoretical value into the detection statistic to obtain the fully self-adaptive PM-GLRT-PHE detection method.

As shown in fig. 1, embodiment 1 of the present invention proposes a space-time adaptive detection method suitable for a partial uniform reverberation environment, where the method includes:

step 1) acquiring a data matrix to be detected and an auxiliary data matrix of echo data received by a spatially symmetric linear array;

step 2) combining the data to be detected and the auxiliary data, and estimating parameters of the pre-constructed detection statistics by adopting maximum likelihood estimation;

and 3) inputting the estimated parameters into a pre-established space-time adaptive detector to finish the adaptive detection of the target.

The method of the present invention will be described in detail below.

1. Description of the problem

Firstly, a multi-channel discrete time signal model of a target and an interference echo is introduced, and on the basis, a binary hypothesis test problem of the target under a target energy leakage sampling model is provided.

1.1 model of received signals

Assuming that a uniform linear array is composed of N array elements, after a series of signal processing is carried out on a received echo signal, a discrete space-time processing N-dimensional echo complex vector z is obtained_iThe received echo vector z to be detected of the ith distance unit_iCan be expressed as

z_i＝s_i+n_i∈C^N×1 (1)

Wherein C represents a complex field, n_iRepresenting the complex Gaussian interference vector, s_iRepresenting a leaked target signal vector, and when energy leakage exists in a target, signal energy can be leaked into a left adjacent distance unit and a right adjacent distance unit of the target, so that a target energy leakage model consisting of three adjacent distance units is obtained, and s_iCan be expressed as:

wherein alpha represents a complex amplitude factor of the received target echo signal and is an unknown determined parameter; chi shape_p(. h) represents a complex ambiguity function of the transmitted signal; epsilon₀∈[0,T_p]Representing the residual time delay, and can measure the severity of the target energy leakage, T_pIs the pulse width; v represents a target normalized space-domain steering vector; f represents the doppler shift introduced by the target and v; l represents the serial number of the sample to be detected.

For convenient expression, the residual time is delayed by epsilon ∈ T by taking the ith distance unit to be detected as the center_p/2,T_p/2]Redefined as:

1.2 hypothesis testing problem

According to the received signal model in 1.2, the binary hypothesis testing problem for leakage target detection can be written as:

wherein H₁And H₀Respectively representing the presence and absence of an object hypothesis. z is a radical of_i∈C^N×1Representing the vector of data to be detected, z_k∈C^{N ×1}The K uniform auxiliary data collected from the distance units adjacent to the data to be detected are represented, and only comprise two interference components of white noise and reverberation. n is_iFor interference components in the data to be detected, under partial uniform reverberation background, with z_kAre independent of each other in statistics, and both are zero-mean complex Gaussian random processes, namely n_i～CN_N(0, γ M) and z_k～CN_N(0, M). γ > 0 represents an unknown energy scaling factor.

In the active sonar system, the covariance matrix M and the normalized airspace guiding vector v of the symmetric linear array have the important characteristic of oblique symmetry, namely

M＝J_NM^*J_N,v＝J_Nv^* (5)

Wherein, J_N∈R^N×NIs a permutation matrix (·)^*For conjugate operations, R is the real number domain, where the permutation matrix J_N∈R^N×NThe matrix is a square matrix with an oblique diagonal of 1 and other elements of 0.

2. Design of detection method

And (5) solving the hypothesis testing problem in the step (4) by adopting a GLRT testing criterion. Suppose Z_L＝[z_l-1,z_l,z_l+1]∈C^N×3For data matrix to be detected, Z_K＝[z₁,...,z_K]∈C^N×KFor the auxiliary data matrix, Z ═ Z_L,Z_K]∈C^N×(3+K)For the joint data matrix, the detection expression based on the GLRT criterion is:

where eta represents a certain false alarm probability (P)_fa) And (4) a lower detection threshold. Based on the skew-symmetric property of M, v, f_j(Z; j ε, j α, γ, M) is represented by_jJ is 0,1 assumes the probability density function of the joint data matrix z:

where det (-) represents the determinant operation of the matrix and tr (-) represents the trace of the matrix, (.)^TAnd (·)^HRepresenting the transpose and conjugate transpose of the matrix,

the fuzzy function matrix D is:

in the formula, t₁,t₂,t₃All are time delays: t is t₁＝-T_p-ε,t₂＝-ε,t₃＝T_p-ε；

Estimation of H using MLE method_jM under the assumption that j is 0,1, the estimation result is obtained as

Substituting (9) into equation (6), the GLRT test decision equation is equivalent to

Based on (10), the MLE of α is equivalent to:

and (3) solving the first-order partial derivative of the alpha according to the formula (11) and setting zero, and obtaining an estimation result of the alpha as follows:

wherein Q is an intermediate variable matrix: denotes S^-1/2v's spanned subspace matrixOrthogonal complement of (I)_NIs an N-dimensional unit matrix;

will be provided withSubstitution (10), solve at H₁Maximum likelihood estimation assuming gammaComprises the following steps:

the intermediate matrices B and C are:

where the matrix Q can be characterized as Q ═ U (γ)₁I₆+Λ)U^HWherein U is equal to C^6×6Is a unitary matrix with lambda being a characteristic value of lambda₁,...,λ₆A diagonal matrix of (c);

the intermediate matrices E and G are:

matrices W and V are:

substituting the characteristic decomposition of Q into (14),

wherein the content of the first and second substances, **

m＝p^*，n＝q^*；

to h₁(γ₁) To obtain gamma₁The first partial derivative of (a) is merged to zero to obtain a pair gamma₁Is estimated value of No form of analytical solution is given and therefore a numerical solution is required, for example a fsolve function may be used to solve the non-linear equation.

Also based on the formula (10) in H⁰Let the MLE for γ be

Is simplified to obtain

To X^HS^-1Performing characteristic decomposition on X to obtainWherein U is₀∈C^6×6Being a unitary matrix, Λ₀For a characteristic value of λ_0,1,...,λ_0,6Diagonal matrix of (2). Substituted into (19) and simplified to obtain

To h₀(γ₀) To obtain gamma₀The first partial derivative of (a) is set to zero, and the pair gamma is obtained as well₀Is estimated value ofHere, theNumerical methods are still required for solving.

And finally, substituting all the obtained unknown parameter estimated values into (14) to derive PM-GLRT-PHE under the partial uniform reverberation environment:

it should be noted that, since the residual delay epsilon estimation has no analytic solution, a grid search method is adopted to estimate the residual delay epsilon. In the end of this process,is reflected in the accuracy of the estimation of the target distance in the range unit to be detected, here by the distance root mean square errorTo indicate.

Detection probability P of PM-GLRT-PHE by adopting Monte Carlo method_dAnd evaluating the target distance estimation performance, and comparing with the existing PM-AMF-PHE, GLRT-LC-PHE and P-GLRT detection methods. Hypothesis false alarm probability P_fa＝10^-4，P_dAnd delta_rmsHas an independent simulation time of 10³N is 12 and γ is 2. Consider the case where the amount of assistance data is limited, i.e., where the amount of assistance data is K — N + 1. Reverberation models typically employ an exponentially-correlated complex gaussian model, i.e., M_i,j＝ρ^|i-j|Where ρ is 0.9 is a hysteresis correlation systemAnd (4) counting. Signal to reverberation noise ratio SRNR |. alpha-²v^HM^-1v/γ。

FIG. 2 shows P of four detection methods_dGraph with SRNR. The simulation result shows that P of the four detection methods_dAll increase with SRNR and P of PM-GLRT-PHE_dIs obviously superior to other existing detection methods of the same type. For example at P_dAt 0.8 PM-GLRT-PHE outperforms PM-AMF-GLRT, GLRT-LC-PHE and P-GLRT with performance gains of about 1.5dB, 4.5dB and 8.5dB, respectively. Similarly, FIG. 3 shows the target distance root mean square error δ for each detection method under the same parameters_rmsBecause P-GLRT does not have a distance estimation capability, it is not evaluated for distance estimation accuracy. As can be seen from the figure, the distance estimation accuracy of PM-GLRT-PHE and PM-AMF-GLRT is approximately the same, and both are better than that of GLRT-LC-PHE. The estimation accuracy of the PM-GLRT-PHE and PM-AMF-PHE detection methods to epsilon is close, and the application of the oblique symmetry characteristic greatly improves the estimation accuracy of epsilon under the condition that auxiliary data is limited.

The inventive method is characterized in that:

1. the invention provides a high-performance space-time self-adaptive detection method suitable for a partial uniform reverberation environment.

2. The invention provides a numerical estimation form of an unknown parameter gamma based on the data to be detected and the auxiliary data, can adopt a numerical solving method such as a fsolve function and the like, and has accurate parameter estimation precision.

3. Aiming at an active sonar system with a space symmetric linear array in a partial uniform reverberation environment, when a received signal is modeled, a target energy leakage sampling model is adopted for a target signal echo to make up for energy leakage loss, and the oblique symmetry characteristic of a covariance matrix is adopted for reverberation to reduce the required quantity of auxiliary data, so that the detection performance when the quantity of the auxiliary data is limited is improved.

4. The present invention assumes that the detected target doppler is known for computing the spatial steering vector.

Embodiment 2 of the present invention provides a space-time adaptive detection system suitable for a partial uniform reverberation environment, where the system includes: the device comprises a pre-established space-time adaptive detector, a data acquisition module, a parameter estimation module and a detection module;

the data acquisition module is used for acquiring a data matrix to be detected and an auxiliary data matrix of echo data received by the spatially symmetric linear arrays;

the parameter estimation module is used for combining the data to be detected and the auxiliary data and estimating the parameters of the pre-constructed detection statistics by adopting maximum likelihood estimation;

and the detection module is used for inputting the estimated parameters into the space-time self-adaptive detector to finish the self-adaptive detection of the target.

Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and are not limited. Although the present invention has been described in detail with reference to the embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the spirit and scope of the invention as defined in the appended claims.