Direction estimation method and device based on composite step length convex combination minimum variance criterion

Abstract

The invention belongs to the technical field of underwater acoustic communication, and relates to a direction estimation method and device based on a composite step length convex combination minimum variance criterion, and the method comprises the steps: setting a control parameter, respectively initializing two filter weight vectors, and initializing a mixed parameter lambda; establishing a mathematical model of an array receiving signal, and determining statistical characteristics of noise; establishing a mathematical model of noise; self-adaptive circulation is carried out, each new data snapshot is processed, weight vectors and mixed parameters of the two filters are continuously updated, and the optimal solution of the weight vector of the weight combination of the two filters is gradually approached; the optimal weight vector obtained through self-adaptive circulation is converted into a final spatial spectrum, and a direction-of-arrival estimation result is displayed; and extracting a final direction-of-arrival estimation value from the spatial spectrum. According to the method, the robustness is improved, the precision and the convergence speed are both considered, the adaptability is high, the engineering realizability is good, and the method is suitable for being popularized and applied to the fields of underwater target positioning, underwater acoustic communication, array signal processing and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of underwater acoustic communication, and particularly relates to a direction estimation method and device based on a composite step convex combination minimum variance criterion. BACKGROUND

[0002] Direction of arrival (DOA) is a key problem in underwater acoustic communication and array signal processing, which involves accurately identifying the incident angles of multiple signal sources in the received array signal. The DOA estimation problem can be essentially regarded as a process of distinguishing the signal subspace from the noise subspace by optimizing the signal model parameters. At present, the mainstream algorithms include the complex total maximum Versoria criterion (CT-MVC) and the variable step size gradient descent total least squares (VSS-GDTLS) algorithm. In view of the significant impulsive interference and non-Gaussian alpha stable noise in the underwater acoustic environment, the traditional fixed step or single filter minimum variance method is limited in performance under this condition, and it is difficult to balance the convergence speed and steady-state accuracy. SUMMARY

[0003] In order to solve the above technical problems, the present application provides a direction estimation method based on a composite step convex combination minimum variance criterion, which adopts the following technical scheme, comprising:

[0004] Setting control parameters, the control parameters including the number of array elements M, the number of signal sources K, the number of sampling shots N, the first filter step size parameter μ1, the second filter step size parameter μ2, the hybrid parameter step size μ λ , the smoothing factor γ, the regularization constant δ, initializing the first filter weight vector and the second filter weight vector respectively w1(0)=w2(0)=0, initializing the hybrid parameter λ λ(0)=0 as the initial value of adaptive weight update;

[0005] Establishing a mathematical model of array received signal x(n)=A(θ)s(n)+v(n), and explicitly defining the statistical characteristics of noise, wherein θ is the direction, x(n) is the received array signal vector at the n th shot time, A(θ) is the steering matrix, a(θ) is the steering vector, s(n) is the source signal vector at the n th shot time, v(n) is the noise vector at the n th shot time, and it obeys the complex alpha stable distribution;

[0006] Establishing a mathematical model of noise v(n)~S(α,β,γ,μ), wherein α is the characteristic exponent, α∈(0,2], β is the skewness parameter, β∈[-1,1], γ is the scale parameter, γ>0, and μ is the location parameter,

[0007] An adaptive loop is performed to process each new data snapshot, and to update the first filter weight vector w1, the second filter weight vector w2 and the mixing parameter β, so as to gradually approach the optimal solution of the combined weight vector of the first filter weight and the second filter weight;

[0008] The optimal weight vector obtained in the adaptive loop is converted into a final spatial spectrum, and a direction of arrival (DOA) estimation result is displayed;

[0009] A final direction of arrival (DOA) estimation value is extracted from the spatial spectrum.

[0010] Preferably, the control parameters are set, the control parameters including the number of array elements M, the number of signal sources K, the number of sampling snapshots N, the first filter step size parameter μ1, the second filter step size parameter μ2, the mixing parameter step size μ λ , the smoothing factor γ, the regularization constant δ, the first filter weight vector w1 and the second filter weight vector w2 are initialized as w1(0)=w2(0)=0, and the mixing parameter λ is initialized as λ(0)=0 as the initial value of the adaptive weight update.

[0011] The number of array elements M, the number of signal sources K, the number of sampling snapshots N, the first filter step size parameter μ1, the second filter step size parameter μ2, the mixing parameter step size μ λ , the smoothing factor γ, and the regularization constant δ are set.

[0012] The first filter weight vector w1 and the second filter weight vector w2 are initialized as w1(0)=w2(0)=0.

[0013] The mixing parameter λ is initialized as λ(0)=0 as the initial value of the adaptive weight update.

[0014] Preferably, the mathematical model of the array received signal x(n)=A(θ)s(n)+v(n) is established, and the statistical characteristics of the noise are specified, where θ is the direction, x(n) is the received array signal vector at the nth snapshot, A(θ) is the steering matrix, a(θ) is the steering vector, s(n) is the source signal vector at the nth snapshot, and v(n) is the noise vector at the nth snapshot, and the step of modeling the steering vector a(θ) of a single far-field narrowband signal specifically includes:

[0015] The steering vector a(θ) of a single far-field narrowband signal is modeled.

[0016] The received signal of the combination of multiple signals and noises is modeled, and a mathematical model of the array received signal is established as x(n)=A(θ)s(n)+v(n), wherein θ is a direction, x(n) is a received array signal vector at the n th snapshot, A(θ) is a steering matrix, a(θ) is a steering vector, s(n) is a source signal vector at the n th snapshot, and v(n) is a noise vector at the n th snapshot, which is subject to a complex α-stable distribution.

[0017] Preferably, the mathematical model of the noise is established as v(n)~S(α,β,γ,μ), wherein α is a characteristic exponent, α∈(0,2], β is a skewness parameter, β∈[-1,1], γ is a scale parameter, γ>0, and μ is a location parameter. The step of establishing the mathematical model of the noise specifically comprises:

[0018] The noise vector v(n) is assumed to be subject to a distribution v(n)~S(α,β,γ,μ), wherein v(n)=[v1(n),v2(n),…,vM-1(n)]T, ~ represents being subject to a distribution, S(α,β,γ,μ) represents a stable distribution determined by four parameters, α is a characteristic exponent, α∈(0,2], β is a skewness parameter, β∈[-1,1], γ is a scale parameter, γ>0, and μ is a location parameter. M T

[0019] Preferably, the step of performing the adaptive loop, processing each new data snapshot, and constantly updating the first filter weight vector w1, the second filter weight vector w2, and the mixing parameter β to gradually approach the optimal solution of the combined weight vector of the first filter weight and the second filter weight specifically comprises:

[0020] The signal of the first element of the array is taken as a reference signal d(n), and the signals of the remaining elements are taken as auxiliary input vectors u(n), d(n)=x1(n), u(n)=[x2(n),x3(n),…,xM-1(n)]T, wherein d(n) is an expected response or a reference signal, and u(n) is an input vector of an adaptive filter with a dimension of (M-1)×1. M T

[0021] The outputs y i (n) and errors e i (n) of the first filter and the second filter are respectively calculated. e i (n) = d(n)-y i (n), i=1,2, y i (n) represents an output signal of the i th filter at time n. ​​​​denotes the conjugate transpose of the weight vector of the i-th filter at time n, e i (n) denotes the instantaneous error signal of the i-th filter at time n;

[0022] The weights of the first filter and the second filter are updated respectively using the normalized least mean square criterion, where * and H denote the conjugate and the conjugate transpose of a complex number respectively, ||u(n)| 2 H (n)u(n) denotes the instantaneous power of the input vector, and δ denotes a regularization constant;

[0023] The mixing parameter λ(n) is adaptively updated to dynamically adjust the combination ratio of the first filter and the second filter, where g'(λ)=g(λ)(1-g(λ)), e a (n)=g(λ(n))e1(n)+(1-g(λ(n)))e2(n), λ(n) denotes an adaptive parameter used to generate the combination coefficient, g(λ(n)) is a Sigmoid function that maps λ(n) to the interval (0, 1) as the combination weight of the first filter, and the weight of the second filter is 1-g(λ(n)), g'(λ(n)) is the derivative of the Sigmoid function, e a (n) is the overall output error of the convex combination, and Re{·} denotes taking the real part of a complex number;

[0024] The weights of the first filter and the second filter are combined according to the current ratio to calculate the combination weight vector w a (n)=g(λ(n))w1(n)+(1-g(λ(n)))w2(n), where w a (n) denotes the combination weight vector at time n;

[0025] The combination weight vector w a is expanded to a complete weight vector where w full is a complete adaptive weight vector with a dimension of Mx1.

[0026] Preferably, the step of converting the optimal weight vector obtained through the adaptive loop into a final spatial spectrum to display the direction of arrival (DOA) estimation result specifically comprises:

[0027] Iterate over all angles θ to perform angle grid scanning;

[0028] For each angle θ, calculate the spatial spectrum where a(θ) is a steering vector corresponding to the scanning angle θ, ∈ is a positive number, and P(θ) is the spatial spectrum power at the angle θ. ​

[0029] Preferably, the step of extracting the final Direction of Arrival (DOA) estimation value from the spatial spectrum comprises the following steps:

[0030] performing peak detection to find all local maximum points, i.e. peaks, on the spatial spectrum P(θ);

[0031] sorting all detected peaks in descending order of amplitude;

[0032] selecting the angles corresponding to the top K peaks as the final Direction of Arrival (DOA) estimation value.

[0033] To solve the above technical problems, the application further provides a Direction of Arrival (DOA) estimation device based on a composite step convex combination minimum variance criterion, which adopts the following technical scheme and comprises:

[0034] an initialization module, configured to set control parameters, including the number of array elements M, the number of signal sources K, the number of sampling snapshots N, the first filter step parameter μ1, the second filter step parameter μ2, the mixed parameter step μ λ , the smoothing factor γ, the regularization constant δ, and initialize the first filter weight vector and the second filter weight vector respectively w1(0)=w2(0)=0, and initialize the mixed parameter λ λ(0)=0 as the initial value of adaptive weight update;

[0035] a signal modeling module, configured to establish a mathematical model of array received signals x(n)=A(θ)s(n)+v(n) and explicitly define the statistical characteristics of noise, wherein θ is a direction, x(n) is a received array signal vector at the nth snapshot, A(θ) is a steering matrix, a(θ) is a steering vector, s(n) is a source signal vector at the nth snapshot, and v(n) is a noise vector at the nth snapshot, which obeys a complex α-stable distribution;

[0036] a noise modeling module, configured to establish a mathematical model of noise v(n)~S(α,β,γ,μ), wherein α is a characteristic exponent, α∈(0,2], β is a skewness parameter, β∈[-1,1], γ is a scale parameter, γ>0, and μ is a location parameter,

[0037] an adaptive loop module, configured to perform adaptive loop to process each new data snapshot, and constantly update the first filter weight vector w1, the second filter weight vector w2 and the mixed parameter λ, so as to gradually approach the optimal solution of the combined weight vector of the first filter weight and the second filter weight;

[0038] a conversion module, configured to convert the optimal weight vector obtained by the adaptive loop into a final spatial spectrum, and display the Direction of Arrival (DOA) estimation result.​

[0039] extracting module, used for extracting a final direction of arrival (DOA) estimation value from the spatial spectrum.

[0040] To solve the above technical problems, the application further provides a computer device, which adopts the technical scheme as follows: a memory and a processor, the memory stores computer readable instructions, and the processor implements the steps of the direction estimation method based on the convex combination minimum variance criterion of the composite step length when executing the computer readable instructions.

[0041] To solve the above technical problems, the application further provides a computer readable storage medium, which adopts the technical scheme as follows: the computer readable storage medium stores computer readable instructions, and the processor implements the steps of the direction estimation method based on the convex combination minimum variance criterion of the composite step length when executing the computer readable instructions.

[0042] Compared with the prior art, the application has the following beneficial effects:

[0043] (1) The robustness is improved: by introducing the double-step length parallel update and the adaptive convex combination mechanism, the influence of impulsive interference and non-Gaussian noise is effectively suppressed, and stable performance can still be maintained in an alpha stable distribution and other strong noise environments.

[0044] (2) The accuracy and the convergence speed are considered: the large-step length component guarantees the fast convergence of the algorithm, the small-step length component provides high steady-state accuracy, and the combination realizes the optimal balance between the accuracy and the convergence speed.

[0045] (3) The adaptability is strong: the convex combination weight is dynamically adjusted according to the current error signal characteristics, can adapt to the changes of different signal-to-noise ratios and noise distributions, and avoids performance degradation.

[0046] (4) The engineering realizability is good: the algorithm structure is simple, the calculation amount is low, and the algorithm is easy to realize in an embedded or real-time underwater acoustic processing system, and is suitable for application in the fields of underwater target positioning, underwater acoustic communication and array signal processing. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the schemes in the application, the drawings needed in the description of the embodiments of the application will be briefly introduced as follows: obviously, the drawings in the following description are some embodiments of the application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of the drawings.

[0048] Figure 1 is a flowchart of an embodiment of the direction estimation method based on the convex combination minimum variance criterion of the composite step length of the application;

[0049] Figure 2 is a comparison chart of weight convergence accuracy of the method of the present application and total least square algorithm based on variable step size gradient descent (VSS-GDTLS) and complex total maximum Versoria criterion (CT-MVC) direction estimation algorithm;

[0050] Figure 3 is a comparison chart of tracking and identification results of two randomly generated channels of the total least square algorithm based on variable step size gradient descent, the complex total maximum Versoria criterion direction estimation algorithm and the method of the present application;

[0051] Figure 4 is a comparison chart of DOA estimation effect of the total least square algorithm based on variable step size gradient descent, the complex total maximum Versoria criterion direction estimation algorithm and the method of the present application;

[0052] Figure 5 is a structural schematic diagram of an embodiment of the direction estimation device based on the composite step size convex combination minimum variance criterion of the present application;

[0053] Figure 6 is a structural schematic diagram of an embodiment of the computer device of the present application. DETAILED DESCRIPTION

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used in the description herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the description herein and the claims of the application and the above description of the drawings herein, the terms "comprising", "comprises" and "having" and any variations thereof are intended to cover without limitation; the description herein and the claims of the application or the above description of the drawings herein, the terms "first", "second" and the like are used to distinguish different objects, not to describe a particular order.

[0055] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. It is expressly understood that the embodiments described herein are merely examples from a potentially infinite number of embodiments that are possible.

[0056] In order to make the technical personnel in the art better understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings.

[0057] It should be noted that the direction estimation method based on the convex combination minimum variance criterion with composite step provided in the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the direction estimation device based on the convex combination minimum variance criterion with composite step is generally arranged in the server / terminal device.

[0058] It should be understood that the number of terminal devices, networks and servers is only illustrative. Any number of terminal devices, networks and servers can be provided according to implementation needs.

[0059] Embodiment one

[0060] Please refer to Figure 1 , a flow chart of one embodiment of the direction estimation method based on the convex combination minimum variance criterion with composite step of the present application is shown. The direction estimation method based on the convex combination minimum variance criterion with composite step comprises the following steps:

[0061] Step S1, setting control parameters, the control parameters comprising an array element number M, a signal source number K, a sampling snapshot number N, a first filter step parameter μ1, a second filter step parameter μ2, a mixing parameter step size μ λ , a smoothing factor γ, a regularization constant δ, initializing a first filter weight vector and a second filter weight vector respectively w1(0)=w2(0)=0, initializing a mixing parameter λ λ(0)=0 as an initial value of adaptive weight update.

[0062] In the embodiment, the electronic device (for example, a server / terminal device) on which the direction estimation method based on the convex combination minimum variance criterion with composite step runs can receive a direction estimation request based on the convex combination minimum variance criterion with composite step through a wired connection mode or a wireless connection mode. It should be noted that the wireless connection mode can include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAXX connection, Zigbee connection, UWB (ultrawideband) connection, and other now known or future developed wireless connection modes.

[0063] In the embodiment, step S1 can specifically comprise the following steps:

[0064] S11, setting an array element number M, a signal source number K, a sampling snapshot number N, a first filter step parameter μ1, a second filter step parameter μ2, a mixing parameter step size μ λ , a smoothing factor γ, a regularization constant δ.

[0065] The array element number M, the signal source number K and the sampling snapshot number N define the physical dimension, the first filter step parameter μ1 and the second filter step parameter μ2 determine the convergence performance, and the mixing parameter step size μ λThe mixing speed is controlled, the smoothing factor γ ensures the power estimation is smooth, the regularization constant δ prevents division by zero errors, and the robustness is enhanced.

[0066] S12, respectively, the first filter weight vector and the second filter weight vector are initialized w1(0)=w2(0)=0.

[0067] The initial weights are provided for both adaptive filters. Starting from the zero-knowledge state, the optimal solution is learned entirely from the subsequent observation data.

[0068] S13, the mixing parameter λ is initialized λ(0)=0 as the initial value of adaptive weight update.

[0069] Through step S13, the control parameter of the convex combination is initialized. Since the Sigmoid function g(0)=0.5, this ensures that the two filters are not biased at the beginning of the algorithm, and they are given equal weights.

[0070] Any adaptive algorithm requires an initial state and control parameters. Suitable parameters are the guarantee of algorithm performance, and zero initialization is a common and stable starting point. By adopting step S1, the repeatability, configurability and numerical stability of the algorithm are provided.

[0071] Step S2, a mathematical model of the array receiving signal x(n)=A(θ)s(n)+v(n) is established, and the statistical characteristics of the noise are determined, wherein θ is the direction, x(n) is the receiving array signal vector at the nth snapshot time, A(θ) is the steering matrix, a(θ) is the steering vector, s(n) is the source signal vector at the nth snapshot time, v(n) is the noise vector at the nth snapshot time, and obeys the complex α stable distribution.

[0072] The noise in the underwater acoustic environment is not Gaussian at all, and adopting a more accurate α stable distribution model is a prerequisite for designing a robust algorithm.

[0073] In the embodiment, step S2 can specifically include the following steps:

[0074] S21, the steering vector a(θ) of a single far-field narrowband signal is modeled.

[0075] The steering vector is constructed, that is, a vector is formed, which completely describes the spatial response characteristics of the array to the signal from the direction θ. The steering vector a(θ) is an intermediary that connects the signal direction and the received data.

[0076] Consider an array of arbitrary geometry consisting of M elements, and assume that a far-field narrowband signal s(t) from the direction θ irradiates on the array.

[0077] Let one reference point in the array be the phase reference, for example, the first element or the physical center of the array. The wave path difference between the signal arriving at the mth element and the reference point is denoted as Δ m .

[0078] The phase delay caused by the wave path difference is: where φ m (θ) is the phase delay (in radian) at the mth element when the signal comes from direction θ, and λ is the wavelength of the signal (in meter). Δ m (θ) is the wave path difference (in meter) between the mth element and the reference point when the signal comes from direction θ. It is a function of the element position and the signal direction θ. Therefore, the complex baseband signal received at the mth element is the source signal multiplied by a complex exponential factor (representing the phase delay): where x m (t) is the complex baseband signal received at the mth element at time t, s(t) is the source signal from direction θ, j is the imaginary unit satisfying j 2 = -1, is the complex exponential factor, which represents a phase delay of radian.

[0079] Stacking the responses of all M elements into a vector, we get the steering vector a(θ) where indicates that it is a complex vector with M rows and 1 column, and M is the total number of elements in the array.

[0080] For a uniform linear array with element spacing d, taking the first element as the reference, the wave path difference Δ m = (m-1)dsin(θ), so its steering vector formula can be simplified as: where d is the distance between adjacent elements in the uniform linear array (in meter), and θ is the angle between the signal incident direction and the array normal direction.

[0081] S22, model the composite received signal of multiple signals and noise, and establish a mathematical model of the array received signal x(n) = A(θ)s(n) + v(n), where θ is the direction, x(n) is the received array signal vector at the nth snapshot, A(θ) is the steering matrix, a(θ) is the steering vector, s(n) is the source signal vector at the nth snapshot, v(n) is the noise vector at the nth snapshot, and it obeys the complex α-stable distribution.

[0082] Integrate multiple signal sources, and combine K signals from different directions {θ1, θ2,..., θ KThe responses of the signals to the array are linearly superimposed. A noise model is introduced to add an additive noise term to make the model more consistent with the actual physical system. The final model is formed to express the whole system as a concise matrix equation, which provides the basis for the subsequent estimation algorithm.

[0083] Suppose there are K far-field narrowband signals s1(n), s2(n),..., s K (n) (n is the snapshot index, representing the time sample point). Each signal s K (n) will produce a response on the array, that is, the signal is multiplied by its corresponding steering vector a(θ K ). According to the superposition of linear systems, the total received signal is the linear superposition of the responses of all K signals.

[0084] In addition, there must be noise (including thermal noise, environmental interference, etc.) in the actual system, which is usually modeled as additive white Gaussian noise. The noise on each array element is denoted as v m (n).

[0085] Therefore, at the n th snapshot, the complete received signal vector x(n) of the array can be expressed as: where x(n) is the array received signal vector at the n th snapshot, k is the index of the signal source, from 1 to K, a(θ k ) is the steering vector corresponding to the direction θ k of the k th signal source, s k (n) is the k th source signal at the n th snapshot, and v(n) is the additive noise vector at the n th snapshot.

[0086] For convenience of calculation, it is rewritten in a compact matrix form: x(n) = A(θ) s(n) + v(n). is the array received signal vector. is the steering matrix, which is composed of K steering vectors side by side, A(θ) = [a(θ1), a(θ2),..., a(θ K )]. is the source signal vector, s(n) = [s1(n), s2(n),..., s K (n)] T , where [·] T represents transposition. is the additive noise vector. It is assumed that each component is a complex Gaussian white noise with mean zero, variance σ 2 , and is independent of the signal, M is the number of array elements, and K is the number of signal sources.

[0087] Step S3, a mathematical model of the noise is established, v(n) ~ S(α, β, γ, μ), wherein α is a characteristic index, α ∈ (0, 2], β is a skewness parameter, β ∈ [-1, 1], γ is a scale parameter, γ > 0, and μ is a location parameter,

[0088] In the embodiment, step S3 can specifically include the following steps:

[0089] S31, a distribution assumption is made on the noise vector v(n), v(n) ~ S(α, β, γ, μ), wherein v(n) = [v1(n), v2(n), …, vM-1(n)]T, and S(α, β, γ, μ) represents a stable distribution determined by four parameters, α is a characteristic index, α ∈ (0, 2], β is a skewness parameter, β ∈ [-1, 1], γ is a scale parameter, γ > 0, and μ is a location parameter, M (n)] T , ~ represents being subjected to a distribution, and S(α, β, γ, μ) represents a stable distribution determined by four parameters, α is a characteristic index, α ∈ (0, 2], β is a skewness parameter, β ∈ [-1, 1], γ is a scale parameter, γ > 0, and μ is a location parameter,

[0090] The characteristic index α determines the degree of pulse of the distribution. The smaller α is, the heavier the tail of the distribution is, and the higher the probability of occurrence of a large amplitude pulse is. The Gaussian distribution is a special case when α = 2. The skewness parameter β represents the degree of asymmetry of the distribution. The scale parameter γ is similar to the standard deviation of the Gaussian distribution, and determines the dispersion degree of the distribution. The location parameter μ is similar to the mean of the Gaussian distribution.

[0091] The traditional algorithm is based on the second-order statistics (variance) and assumes that the noise is Gaussian. However, in actual environments such as underwater acoustic and radar, the noise often has significant pulse characteristics, and the variance can be infinite, which makes the traditional algorithm invalid. The α-stable distribution model can more accurately describe such pulse noise, thereby guiding the design of more robust algorithms (such as the minimum variance criterion and the normalized update in the present algorithm).

[0092] Step S4, an adaptive loop is performed to process each new data snapshot, and the first filter weight vector w1, the second filter weight vector w2, and the mixing parameter β are updated constantly, so as to gradually approach the optimal solution of the combined weight vector of the first filter weight and the second filter weight.

[0093] In the embodiment, step S4 can specifically include the following steps:

[0094] S41, the signal of the first array element is taken as a reference signal d(n), and the signals of the remaining array elements are taken as an auxiliary input vector u(n), d(n) = x1(n), u(n) = [x2(n), x3(n), …, xM-1(n)]T, M (n)] T , wherein d(n) is an expected response or a reference signal, and u(n) is an input vector of an adaptive filter, and the dimension is (M-1) × 1.

[0095] This operation converts the multi-channel DOA estimation problem into an adaptive filtering problem. The first sensor is taken as reference, and the goal of the filter is to estimate the relationship (i.e. weights w) between the signals of other sensors and the reference signal, which contains the spatial information of signals from different directions. This structure is called beamforming-side lobe cancellation structure.

[0096] By step S41, the complex spatial spectrum estimation problem is converted into an adaptive filtering problem, and an adaptive algorithm can be used. The dimension of the input vector u(n) is M-1, which is reduced in dimension, reducing the computational complexity.

[0097] S42, respectively, calculate the output y i (n) and the error e i (n) of the first filter and the second filter, e i (n) = d(n)-y i (n), i = 1, 2, y i (n) represents the output signal of the i-th filter at time n, represents the conjugate transpose of the weight vector of the i-th filter at time n, e i (n) represents the instantaneous error signal of the i-th filter at time n.

[0098] The filter output y i (n) is a prediction of the desired signal d(n). The error signal e i (n) measures the accuracy of the prediction. In the ideal case, if the weights are optimal, the error signal should mainly contain noise components. This error signal is the driving force for updating the filter weights.

[0099] Through step S42, the performance indication can be performed: the size of the error signal directly reflects the good or bad of the current performance of the filter; the update can be driven: the necessary input is provided for the subsequent weight update formula.

[0100] S43, using the normalized least mean square criterion, respectively updating the weights of the first filter and the second filter, where * and H represent the conjugate and conjugate transpose of complex numbers respectively, ||u(n)| 2 = u H (n)u(n) represents the instantaneous power of the input vector, and δ represents the regularization constant.

[0101] The core idea of step S43 is the steepest descent method, and the weights are updated along the negative gradient direction of the instantaneous error square performance surface. Normalization (divided by ||u(n)| 2) to make the update step size insensitive to the power variation of the input signal, which improves the stability and convergence speed of the algorithm. Different step sizes μ1 (larger) and μ2 (smaller) are implemented to make the first filter fast in tracking signal changes but with large steady-state error, and the second filter slow in convergence but with high steady-state accuracy. This is the embodiment of the compound step size.

[0102] Step S43 is adopted to improve robustness: the normalization operation makes the algorithm stable when the amplitude of the input signal varies greatly; parallel double paths are implemented: both fast convergence and high-precision steady state are obtained, providing a basis for subsequent convex combination.

[0103] S44, adaptively update the mixing parameter λ(n), dynamically adjust the combination ratio of the first filter and the second filter, wherein, g'(λ) = g(λ)(1-g(λ)), e a (n) = g(λ(n))e1(n) + (1-g(λ(n)))e2(n), λ(n) represents an adaptive parameter used to generate the combination coefficient, g(λ(n)) is a Sigmoid function that maps λ(n) to the interval (0, 1) as the combination weight of the first filter, the weight of the second filter is 1-g(λ(n)), g'(λ(n)) is the derivative of the Sigmoid function, e a (n) is the overall output error of the convex combination, and Re{·} represents the real part of the complex number.

[0104] The combined instantaneous error power |e a (n)| 2 is minimized by stochastic gradient descent.

[0105] The update amount of λ(n) is proportional to the correlation between the combination error e a (n) and the difference between the outputs of the two filters (y1(n)-y2(n)). If the error of the large step size filter (output y1) is smaller, λ will increase, making its weight g(λ) increase; otherwise, the weight of the small step size filter will be increased. The Sigmoid function can ensure that the combination weight always changes smoothly between (0, 1).

[0106] By adopting step S44, the adaptive algorithm can automatically determine the current environment (convergence initial stage or steady state, high noise or low noise) and select the most suitable filter as the dominant one; the performance can be improved, and the performance of the combination system will not be worse than the best one of the two filters, and is usually better than any single filter.

[0107] S45, combine the weights of the first filter and the second filter according to the current ratio to calculate the combination weight vector w a(n) = g(λ(n))w1(n) + (1 - g(λ(n)))w2(n), where w a (n) represents the combined weight vector at time n.

[0108] The optimal weight vector estimation at current time is generated. The weight vector absorbs the advantages of fast tracking of large step size filter and combines the advantages of high precision of small step size filter.

[0109] Through step S45, the optimal balance is achieved: the real-time optimal balance between convergence speed and steady-state accuracy is achieved; the smooth transition is performed: the Sigmoid function ensures that the change of the combined weight is smooth, avoiding the jump of performance.

[0110] S46, the combined weight vector w a is extended to the complete weight vector w full , where w full is the complete adaptive weight vector, with a dimension of Mx1.

[0111] This operation is the inverse process of step S41. In step S41, the problem is processed by reducing the dimension from M channels to M-1 channels. Now the optimal solution w a obtained (which has the physical meaning of the optimal linear estimation coefficient of other elements to the reference element) needs to be restored to the weight vector of the full array. The first element of the weight vector w full is fixed as 1, corresponding to the reference channel itself.

[0112] By restoring the structure, the complete array weight which can be used to calculate the spatial spectrum is obtained. In line with the convention, the Capon spectrum estimation method needs the complete weight vector.

[0113] Step S5 converts the optimal weight vector obtained by the adaptive loop into the final spatial spectrum, and displays the direction of arrival DOA estimation result.

[0114] In this embodiment, step S5 can specifically include the following steps:

[0115] S51, traverse all angles θ, and perform angle grid scanning.

[0116] Traverse all possible angles θ, for example, 0°:180°, and systematically calculate the power output in each candidate direction.

[0117] S52, for each angle θ, calculate the spatial spectrum where a(θ) is the steering vector corresponding to the scanning angle θ, ∈ is a positive number, and P(θ) is the spatial spectrum power at the angle θ.

[0118] P(θ) is the power output of adaptive beamforming, and its principle is: let the weight vector wfull All possible directions are matched. When the scanning angle θ is equal to the direction of a real signal source, The value will be relatively large (trying to let the signal pass without distortion), resulting in a peak of P(θ). If θ is not the signal direction, w full will form a null in this direction, so that The value is very small, so the output of P(θ) will be large. Therefore, the corresponding signal source direction is the spectral peak.

[0119] ∈ is a safety factor such as 10 -10 , to prevent the denominator from being zero and ensure numerical stability.

[0120] Step S6, extracting the final wave arrival DOA direction estimation value from the spatial spectrum.

[0121] In this embodiment, step S6 can specifically include the steps of:

[0122] S61, performing peak detection to find all local maximum points, i.e. peaks, on the spatial spectrum P(θ).

[0123] The function of step S61 is to locate all possible signal source directions.

[0124] S62, sorting all detected peaks in descending order of amplitude.

[0125] The function of step S62 is to prepare for selecting the most possible K signal sources.

[0126] S63, selecting the angles corresponding to the first K largest peaks as the final wave arrival direction DOA estimation value.

[0127] The number K of signal sources is known or estimated in advance, and the algorithm finds the most possible K wave arrival directions accordingly.

[0128] Figure 2 is the comparison chart of the weight convergence accuracy of the method of the present application and the total least square algorithm based on variable step size gradient descent (VSS-GDTLS) and the complex domain total maximum Versoria criterion (CT-MVC) direction estimation algorithm. As shown in Figure 2 , 10000 points of random Gaussian signals are generated according to the standard normal distribution, the Gaussian noise signal-to-noise ratio is set to 15dB, the impulse interference variance is set to 100 times the variance of the received signal, and α-stable distribution noise is used, wherein the probability of generating impulse interference is Pr=0.25, and the obtained results are as shown in Figure 2As shown, unlike the traditional VSS-GDTLS, CT-MVC method, the application adopts a double-step parallel update and adaptive convex combination mechanism, which can effectively reduce the steady-state error of weight estimation under different received signal-to-noise ratios and different probability impulsive interference, and obtain more accurate weight estimation results.

[0129] Figure 3 The total least square algorithm based on variable step size gradient descent, the complex Versoria criterion direction estimation algorithm and the method of the application are compared in tracking and identification results of two randomly generated channels. Figure 3 As shown, the data length is set to 10000 points, and at the moment when the algorithm iteration is performed to 5000 points, the weight coefficient is suddenly changed, so as to evaluate the tracking ability of each algorithm to the weight coefficient in the time-varying case.

[0130] Figure 4 The total least square algorithm based on variable step size gradient descent, the complex Versoria criterion direction estimation algorithm and the method of the application are compared in tracking and identification results of two randomly generated channels. Figure 4 As shown, in order to further investigate the DOA estimation effect of different algorithms under long fast shot number and alpha stable noise interference, the real angle is set to: [-25, 20, 45], the fast shot number is set to 5000, 200 Monte Carlo experiments are performed on three algorithms to eliminate the interference of uncertainty, and the spectral function images of three algorithms are as shown in Figure 4 As shown, it can be seen that the application is obviously superior to the VSS-GDTLS and CT-MVC methods in the DOA estimation effect, and under 200 Monte Carlo experiments, the average estimation values of three algorithms are: VSS-GDTLS average estimation angle: [-25.475, 12.565, 37.5575], CT-MVC average estimation angle: [-31.2725, 16.5675, 44.43], and CVX-CMVC average estimation angle: [-30.1475, 19.4375, 44.3825].

[0131] The embodiment is implemented, and the beneficial effects are:

[0132] (1) The robustness is improved: by introducing the double-step parallel update and adaptive convex combination mechanism, the influence of impulsive interference and non-Gaussian noise is effectively suppressed, and the stable performance can be maintained in the alpha stable distribution and other strong noise environments.

[0133] (2) Precision and convergence speed are considered: the large step component ensures the fast convergence of the algorithm, the small step component provides higher steady-state accuracy, and the combination realizes the optimal balance of precision and convergence speed.

[0134] (3) Strong adaptability: the convex combination weight is dynamically adjusted according to the current error signal characteristics, which can adapt to the changes of different signal-to-noise ratios and noise distribution, and avoid performance degradation.

[0135] (4) Good engineering realizability: the algorithm structure is simple, the calculation amount is low, and it is easy to realize in embedded or real-time underwater acoustic processing system, and is suitable for popularization and application in underwater target positioning, underwater acoustic communication and array signal processing fields.

[0136] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems,

[0137] microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0138] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by computer-readable instructions instructing related hardware, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments. The storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0139] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0140] Example 2

[0141] Further reference Figure 5 As a response to the above Figure 1 The present invention provides an embodiment of a direction estimation device based on a composite step-size convex combination minimum variance criterion, which is similar to the method described above. Figure 5 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0142] like Figure 5 As shown, the direction estimation device 70 based on the minimum variance criterion of composite step-size convex combination described in this embodiment includes: an initialization module 71, a signal modeling module 72, a noise modeling module 73, an adaptive loop module 74, a conversion module 75, and an extraction module 76. Wherein:

[0143] Initialization module 71 is used to set control parameters, including the number of array elements M, the number of signal sources K, the number of sampling snapshots N, the first filter step size parameter μ1, the second filter step size parameter μ2, and the mixing parameter step size μ. λ Smoothing factor γ, regularization constant δ, initialize the first filter weight vector and the second filter weight vector w1(0)=w2(0)=0 respectively, and initialize the mixing parameter λ(0)=0 as the initial value for adaptive weight update;

[0144] The signal modeling module 72 is used to establish the mathematical model of the array receiving signal x(n)=A(θ)s(n)+v(n) and to define the statistical characteristics of the noise, where θ is the direction, x(n) is the received array signal vector at the nth snapshot, A(θ) is the steering matrix, a(θ) is the steering vector, s(n) is the source signal vector at the nth snapshot, and v(n) is the noise vector at the nth snapshot, which follows a complex α-stable distribution.

[0145] The noise modeling module 73 is used to establish a mathematical model of noise, v(n)~S(α,β,γ,μ), where α is the characteristic exponent, α∈(0,2], β is the skew parameter, β∈[-1,1], γ is the scale parameter, γ>0, and μ is the location parameter.

[0146] The adaptive loop module 74 is used to perform adaptive looping, process each new data snapshot, continuously update the first filter weight vector w1, the second filter weight vector w2 and the mixing parameter β, and gradually approach the optimal solution of the combined weight vector of the first filter weight and the second filter weight.

[0147] Transformation module 75 is used to convert the optimal weight vector obtained by the adaptive loop into the final spatial spectrum, displaying the direction of arrival (DOA) estimation results.

[0148] Extraction module 76 is used to extract the final DOA (Direction of Arrival) estimate from the spatial spectrum.

[0149] The beneficial effects of implementing this embodiment are:

[0150] (1) Improved robustness: By introducing a dual-step parallel update and adaptive convex combination mechanism, the impact of impulse interference and non-Gaussian noise is effectively suppressed, and the stable performance can still be maintained under strong noise environment such as α stable distribution.

[0151] (2) It balances accuracy and convergence speed: the large step size component ensures the fast convergence of the algorithm, while the small step size component provides high steady-state accuracy. The combination achieves the optimal balance between accuracy and convergence speed.

[0152] (3) Strong adaptability: The convex combination weights are dynamically adjusted according to the characteristics of the current error signal, which can adapt to changes in different signal-to-noise ratios and noise distributions and avoid performance degradation.

[0153] (4) Good engineering feasibility: The algorithm has a simple structure and low computational cost, making it easy to implement in embedded or real-time underwater acoustic processing systems. It is suitable for application in fields such as underwater target positioning, underwater acoustic communication and array signal processing.

[0154] Example 3

[0155] To address the aforementioned technical problems, embodiments of the present invention also provide a computer device. Please refer to [link / reference needed]. Figure 6 , Figure 6 This is a basic structural block diagram of the computer device in this embodiment.

[0156] The computer device 8 includes a memory 81, a processor 82, and a network interface 83, which are communicatively connected via a system bus. It should be noted that the computer device 8 is shown with the components memory 81, processor 82, and network interface 83, but it should be understood that not all of the illustrated components are required, and that more or fewer components can be implemented. As will be appreciated by one skilled in the art, the computer device is a device capable of automatically processing numerical and / or information according to pre-set or stored instructions, and its hardware includes, but is not limited to, a microprocessor, an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, and the like.

[0157] The computer device can be a desktop computer, a notebook computer, a palm computer, a cloud server, or the like. The computer device can interact with a user through a keyboard, a mouse, a remote controller, a touchpad, a voice control device, or the like.

[0158] The memory 81 includes at least one type of readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, or the like), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, or the like. In some embodiments, the memory 81 can be an internal storage unit of the computer device 8, such as a hard disk or a memory of the computer device 8. In other embodiments, the memory 81 can also be an external storage device of the computer device 8, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, or the like. Of course, the memory 81 can include both an internal storage unit and an external storage device of the computer device 8. In this embodiment, the memory 81 is generally used to store an operating system and various application software installed in the computer device 8, such as computer readable instructions of the method for estimating a direction based on a minimum variance criterion of a convex combination of steps, and the like. In addition, the memory 81 can also be used to temporarily store various data that has been output or will be output.

[0159] The processor 82 in some embodiments can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 82 is generally used to control the overall operation of the computer device 8. In the present embodiment, the processor 82 is configured to run computer-readable instructions stored in the memory 81 or process data, such as computer-readable instructions of the direction estimation method based on the convex combination minimum variance criterion with composite step size.

[0160] The network interface 83 can include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 8 and other electronic devices.

[0161] The present embodiment has the following beneficial effects:

[0162] (1) Improved robustness: By introducing a double-step-size parallel update and an adaptive convex combination mechanism, the influence of impulsive interference and non-Gaussian noise is effectively suppressed, and stable performance can still be maintained in an alpha-stable distribution environment with strong noise.

[0163] (2) Balancing accuracy and convergence speed: The large-step-size component ensures fast convergence of the algorithm, and the small-step-size component provides high steady-state accuracy, and the combination achieves an optimal balance between accuracy and convergence speed.

[0164] (3) Strong adaptability: The convex combination weight is dynamically adjusted according to the current error signal characteristics, and can adapt to changes in different signal-to-noise ratios and noise distributions, avoiding performance degradation.

[0165] (4) Good engineering realizability: The algorithm structure is simple and has low computational complexity, and is easy to implement in embedded or real-time underwater acoustic processing systems, and is suitable for application in underwater target positioning, underwater acoustic communication, and array signal processing fields.

[0166] Embodiment Four

[0167] The present application also provides another embodiment, that is, a computer readable storage medium storing computer readable instructions, the computer readable instructions being executable by at least one processor to cause the at least one processor to perform the steps of the direction estimation method based on the convex combination minimum variance criterion with composite step size as described above.

[0168] The present embodiment has the following beneficial effects:

[0169] (1) Improved robustness: By introducing a double-step-size parallel update and an adaptive convex combination mechanism, the influence of impulsive interference and non-Gaussian noise is effectively suppressed, and stable performance can still be maintained in an alpha-stable distribution environment with strong noise.

[0170] (2) Precision and convergence speed are considered: the large step component ensures the fast convergence of the algorithm, and the small step component provides higher steady-state precision, and the combination realizes the optimal balance of precision and convergence speed.

[0171] (3) Strong adaptability: the convex combination weight is dynamically adjusted according to the current error signal characteristics, which can adapt to the changes of different signal-to-noise ratios and noise distribution, and avoid performance degradation.

[0172] (4) Good engineering realizability: the algorithm structure is simple and the calculation amount is low, which is easy to realize in embedded or real-time underwater acoustic processing system, and is suitable for application in underwater target positioning, underwater acoustic communication and array signal processing fields.

[0173] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by means of software and the necessary general hardware platform, of course, they can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the various embodiment methods of the present application.

[0174] Obviously, the above-described embodiments are only some of the embodiments of the present application, not all the embodiments, and the preferred embodiments of the present application are given in the drawings, but do not limit the patent scope of the present application. The present application can be realized in many different forms, and on the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing specific embodiments, or make equivalent replacements to some technical features. Any equivalent structure made by using the content of the present application specification and drawings, directly or indirectly applied to other related technical fields, is also within the patent protection scope of the present application.

Claims

1. A method for direction estimation based on a composite step convex combination minimum variance criterion, characterized in that, The method comprises the following steps: Setting control parameters, the control parameters including the number of array elements M, the number of signal sources K, the number of sampling shots N, the first filter step parameter μ1, the second filter step parameter μ2, the mixing parameter step μ λ , the smoothing factor γ, the regularization constant δ, initializing the first filter weight vector and the second filter weight vector respectively w1(0) = w2(0) = 0, initializing the mixing parameter λ λ(0) = 0 as the initial value of adaptive weight update; a mathematical model of array receiving signals x(n)=A(θ)s(n)+v(n) is established, and statistical characteristics of noise are determined, wherein θ is a direction, x(n) is a receiving array signal vector at the n-th snapshot, A(θ) is a steering matrix, a(θ) is a steering vector, s(n) is a source signal vector at the n-th snapshot, and v(n) is a noise vector at the n-th snapshot, and the noise vector obeys a complex α stable distribution; A mathematical model of noise is established v(n) ~ S(a, β, γ, μ), wherein a is a characteristic exponent, a e (0, 2], β is a skewness parameter, β e [-1, 1], γ is a scale parameter, γ > 0, and μ is a location parameter, an adaptive loop is performed, each new data snapshot is processed, the first filter weight vector w1, the second filter weight vector w2 and the mixing parameter β are updated constantly, and an optimal solution of a combined weight vector of the first filter weight and the second filter weight is gradually approached; the optimal weight vector obtained through the adaptive loop is converted into a final spatial spectrum, and a direction of arrival (DOA) estimation result is displayed; a final direction of arrival (DOA) estimation value is extracted from the spatial spectrum.

2. The method of claim 1, wherein the method is based on a composite step-size convex combination minimum variance criterion. The setting control parameters, the control parameters include array element number M, signal source number K, sampling snapshot number N, first filter step parameter μ1, second filter step parameter μ2, mixing parameter step μ λ , smoothing factor γ, regularization constant δ, respectively, the first filter weight vector and the second filter weight vector are initialized w1(0) = w2(0) = 0, the mixing parameter λ is initialized λ(0) = 0, as the initial value of adaptive weight update step specifically includes: Setting the number of array elements M, the number of signal sources K, the number of sampling shots N, the first filter step size parameter μ1, the second filter step size parameter μ2, the mixing parameter step size μ λ , the smoothing factor γ, and the regularization constant δ The first filter weight vector and the second filter weight vector are initialized respectively as w1(0)=w2(0)=0; the mixing parameter λ is initialized as λ(0)=0, and serves as an initial value of adaptive weight updating.

3. The method of claim 1, wherein the method is based on a composite step-size convex combination minimum variance criterion. The step of establishing the mathematical model of array receiving signals x(n)=A(θ)s(n)+v(n) and determining statistical characteristics of noise, wherein θ is a direction, x(n) is a receiving array signal vector at the n-th snapshot, A(θ) is a steering matrix, a(θ) is a steering vector, s(n) is a source signal vector at the n-th snapshot, and v(n) is a noise vector at the n-th snapshot, and the noise vector obeys a complex α stable distribution, specifically comprises the following steps: a steering vector a(θ) of a single far-field narrow-band signal is modeled; a combined receiving signal of multiple signals and noise is modeled, and the mathematical model of array receiving signals x(n)=A(θ)s(n)+v(n) is established, wherein θ is a direction, x(n) is a receiving array signal vector at the n-th snapshot, A(θ) is a steering matrix, a(θ) is a steering vector, s(n) is a source signal vector at the n-th snapshot, and v(n) is a noise vector at the n-th snapshot, and the noise vector obeys a complex α stable distribution.

4. The method of claim 3, wherein, The step of establishing the mathematical model of noise v(n)~S(a, b, g, m), wherein a is a characteristic index, a e (0, 2], b is a skewness parameter, b e [-1, 1], g is a scale parameter, g > 0, and m is a location parameter, specifically comprises: The noise vector v(n) is assumed to be distributed as v(n) ~ S(a, β, γ, μ), where v(n) = [v1(n), v2(n),..., vn(n)]T, and ~ denotes to be subject to a distribution, S(a, β, γ, μ) denotes a stable distribution determined by four parameters, a is a characteristic exponent, a e (0, 2], β is a skewness parameter, β e [-1, 1], γ is a scale parameter, γ > 0, and μ is a location parameter, M (n)] T , ~ denotes to be subject to a distribution, S(a, β, γ, μ) denotes a stable distribution determined by four parameters, a is a characteristic exponent, a e (0, 2], β is a skewness parameter, β e [-1, 1], γ is a scale parameter, γ > 0, and μ is a location parameter, 5. The method of claim 4, wherein, The step of performing the adaptive loop, processing each new data snapshot, constantly updating the first filter weight vector w1, the second filter weight vector w2 and the mixing parameter β, and gradually approaching an optimal solution of a combined weight vector of the first filter weight and the second filter weight, specifically comprises the following steps: Take the signal of the first array element as the reference signal d(n), and the signals of the remaining array elements as the auxiliary input vector u(n), d(n) = x1(n), u(n) = [x2(n), x3(n),..., xn(n)]T M (n)] T where d(n) is the desired response or reference signal, and u(n) is the input vector of the adaptive filter, with a dimension of (M-1) x 1; calculating the output y of the first filter and the output y i (n) and the error e i (n), e i (n) = d(n) - y i (n), i = 1, 2, y i (n) denotes the output signal of the i-th filter at time n, denotes the conjugate transpose of the weight vector of the i-th filter at time n, e i (n) denotes the instantaneous error signal of the i-th filter at time n; The weights of the first filter and the second filter are updated respectively using a normalized least mean square criterion, where * and H represent the conjugate and the conjugate transpose of complex numbers respectively, ||u(n)| 2 = u H (n)u(n) represents the instantaneous power of the input vector, and δ represents a regularization constant; adaptively updating the mixing parameter λ(n) to dynamically adjust the combined proportion of the first filter and the second filter, wherein, g'(λ) = g(λ)(1 - g(λ)), e a (n) = g(λ(n))e1(n) + (1 - g(λ(n)))e2(n), λ(n) represents an adaptive parameter used to generate the combination coefficient, g(λ(n)) is a Sigmoid function that maps λ(n) to the interval (0, 1) as the combination weight of the first filter, the weight of the second filter is 1 - g(λ(n)), g'(λ(n)) is the derivative of the Sigmoid function, e a (n) is the overall output error of the convex combination, and Re{·} represents taking the real part of a complex number; combining the weights of the first filter and the second filter according to the current proportion to calculate a combined weight vector w a (n) = g(λ(n))w1(n) + (1 - g(λ(n)))w2(n), where w a (n) represents the combined weight vector at time n; The combined weight vector w a is extended to the full weight vector w full , where w full is the full adaptive weight vector of dimension M x 1.

6. The direction estimation method based on the minimum variance criterion of composite step-size convex combination according to claim 5, characterized in that, The step of converting the optimal weight vector obtained through the adaptive loop into a final spatial spectrum and displaying a direction of arrival (DOA) estimation result, specifically comprises the following steps: all angles θ are traversed, and angle grid scanning is performed; For each angle θ, the spatial spectrum is calculated where a(θ) is the steering vector corresponding to the scan angle θ, ∈ is a positive number, and P(θ) is the spatial spectrum power at angle θ.

7. The direction estimation method based on the minimum variance criterion of composite step-size convex combination according to claim 6, characterized in that, The step of extracting a final direction of arrival (DOA) estimation value from the spatial spectrum, specifically comprises the following steps: peak value detection is performed, and all local maximum points, that is, peak values, on the spatial spectrum P(θ) are searched; all detected peak values are sorted in descending order of amplitude; selecting the angles corresponding to the K largest peak values as the final direction of arrival, DOA, estimate.

8. A device for direction estimation based on a composite step convex combination minimum variance criterion, characterized in that, The method comprises the following steps: An initialization module is configured to set control parameters, including the number of array elements M, the number of signal sources K, the number of sampling shots N, the first filter step parameter μ1, the second filter step parameter μ2, the mixing parameter step μ λ , the smoothing factor γ, the regularization constant δ, and initialize the first filter weight vector and the second filter weight vector respectively w1(0) = w2(0) = 0, and initialize the mixing parameter λ λ(0) = 0 as the initial value of adaptive weight update. A signal modeling module is configured to establish a mathematical model of the array receiving signal x(n) = A(θ)s(n) + v(n) and to explicitly define statistical characteristics of the noise, wherein θ is a direction, x(n) is a receiving array signal vector at the nth snapshot, A(θ) is a steering matrix, a(θ) is a steering vector, s(n) is a source signal vector at the nth snapshot, v(n) is a noise vector at the nth snapshot, and the noise is subject to a complex α stable distribution. a noise modeling module for establishing a mathematical model of the noise v(n) ~ S(a, b, g, m), wherein a is a characteristic exponent, a e (0, 2], b is a skewness parameter, b e [-1, 1], g is a scale parameter, g > 0, and m is a location parameter, An adaptive loop module is configured to perform an adaptive loop, to process each new data snapshot, and to constantly update the first filter weight vector w1, the second filter weight vector w2, and the mixing parameter β, so as to gradually approach an optimal solution of the combined weight vector of the first filter weight and the second filter weight. A conversion module is configured to convert the optimal weight vector obtained by the adaptive loop into a final spatial spectrum, and to display a direction of arrival (DOA) estimation result. An extraction module is configured to extract a final DOA direction estimation value from the spatial spectrum.

9. A computer device, comprising: A memory and a processor are included, the memory has computer readable instructions stored therein, and the processor, when executing the computer readable instructions, implements the steps of the direction estimation method based on the convex combination minimum variance criterion of a composite step length according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium has computer readable instructions stored thereon, and the computer readable instructions, when executed by a processor, implement the steps of the direction estimation method based on the convex combination minimum variance criterion of a composite step length according to any one of claims 1 to 7.

Images