A DOA estimation algorithm for wideband signals in 6G ISAC system
By introducing virtual nodes and GAMP algorithms in 6G ISAC systems, based on the probability model of Dilicre process prior, the problem of poor performance and high complexity of broadband DOA estimation technology under unknown conditions is solved, and high precision and low complexity are achieved DOA estimate of broadband signal.
Patent Information
- Application Number
- CN202311181504.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-13
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2043-09-13
AI Technical Summary
The existing broadband DOA estimation technology has poor performance under unknown number of sources, few snapshots and highly correlated angles. The method based on the sparse Bayesian learning framework assumes that the subband signals share the same frequency band, which does not conform to the actual situation. This makes it difficult to take into account both estimation accuracy and complexity.
A virtual node was introduced, combined with the GAMP algorithm, and a DP-GAMP algorithm was proposed. A sparse signal model was constructed based on the probability model of Dirichlet's process prior, and iteratively updated to improve the DOA estimation accuracy and reduce the algorithm complexity.
High-precision broadband signal DOA estimation in 6G ISAC systems is realized, while reducing algorithm complexity and improving estimation efficiency.
Smart Images

Figure CN117388791B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of communication broadband technology, and more specifically, to a 6G ISAC system broadband signal DOA estimation algorithm. Background Art
[0002] The 6G communication network will integrate the digital world and the physical world. It is not only a communication transmission channel, but also can sense everything and realize the intelligence of everything. 6G will become a network of sensors and machine learning, with the data center as the brain and machine learning throughout the entire network. Perception-assisted communication methods such as location-based beamforming and tracking can greatly improve communication performance. The integrated interawareness system (ISAC) in 6G can be considered a key technology to promote the integration of the physical, biological and digital worlds in the future. DOA estimation is an important research direction in the field of array signal processing. It is widely used in exploration, positioning, channel estimation and other fields, involving various aspects such as production, life, and military. With the development of 6G technology, the broadband DOA estimation technology has attracted the attention of more and more scholars because it can locate user equipment.
[0003] The traditional method of broadband DOA estimation is to extend the narrowband DOA estimation method to broadband signals, such as Fourier transform method and subspace method. However, these methods perform poorly in the case of unknown number of sources, small number of snapshots and high angle correlation. The DOA estimation algorithm based on the sparse Bayesian learning (SBL) framework can estimate the DOA angle information well, but the above methods all assume that subband signals from different directions share the same frequency band, which is not in line with the actual situation.
[0004] Therefore, a probability model based on Dirichlet process prior is proposed, and the DOA estimation algorithm derived by the mean field algorithm has a high complexity. How to further improve the accuracy of the DOA estimation algorithm while reducing the complexity of the algorithm is an urgent problem to be solved. Summary of the invention
[0005] In order to solve the above problems, the present invention provides a 6G ISAC system broadband signal DOA estimation algorithm, and introduces the GAMP algorithm by introducing virtual nodes, thereby improving the estimation accuracy of the DOA estimation algorithm while reducing the complexity of the algorithm.
[0006] The technical problem to be solved by the present invention is achieved by the following technical solution: A 6G ISAC system broadband signal DOA estimation algorithm comprises the following steps:
[0007] S1: Use M sensors to build a uniform linear array, receive the broadband signals transmitted by K far-field antennas, and divide the K broadband signals into F narrowband signals, where K is a positive integer, to form the output signal matrix Y;
[0008] Based on the angle domain, the range [0°, 180°] is evenly divided into N grids to obtain the angle grid point set Θ = [θ1,…,θ N ], and then define the broadband signal sent by K antennas as an N×F dimensional sparse signal matrix X in the angle domain, define an M×F dimensional noise signal matrix W, and define an M×N dimensional overcomplete dictionary matrix A;
[0009] A sparse signal model of the output signal matrix Y is constructed based on the sparse signal matrix X, the noise signal matrix W and the overcomplete dictionary matrix A;
[0010] S2: Based on the constructed sparse signal model of the output signal matrix Y, a probabilistic broadband signal model based on the Dirichlet process prior is established;
[0011] S3: For each subband, introduce a dummy variable node;
[0012] S4: A variational Bayesian broadband DOA estimation method based on the GAMP algorithm is proposed, namely the DP-GAMP algorithm;
[0013] S5: Initialize the unknown parameters, input the received output signal matrix Y, iteratively update through the DP-GAMP algorithm, and obtain the estimated values of the probability model parameters;
[0014] S6: Calculate the estimated value of DOA based on the obtained parameter estimated values.
[0015] The present invention is further configured as follows: Step S1 comprises:
[0016] M sensors form a uniform linear array with a spacing of Δd to receive broadband signals transmitted by K far-field antennas;
[0017] The broadband signal is divided into F narrowband signals through a filter to form an M×F dimensional output signal matrix Y;
[0018] Then, based on the angle domain, the range [0°, 180°] is evenly divided into N grids to obtain the angle grid point set Θ = [θ1,…,θ N ];
[0019] Then define the broadband signal sent by K antennas as N×F dimensional sparse signal matrix X in the angle domain, define M×F dimensional noise signal matrix W, and define M×N dimensional overcomplete dictionary matrix A;
[0020] Based on the sparse signal matrix X, the noise signal matrix W and the overcomplete dictionary matrix A, the sparse signal model of the output signal matrix Y is constructed as follows:
[0021] Y=AX+W
[0022] in,
[0023] A=[a(θ1),…,a(θ i ),…,a(θ N )],
[0024]
[0025] In the formula, a(θ i ), i=1,…,N means the incident angle is θ i is the guiding vector at , and λ is the wavelength.
[0026] The present invention is further configured as follows: the probabilistic broadband signal model based on the Dirichlet process prior includes:
[0027] Assuming that the additive noise of the broadband signal obeys a complex Gaussian distribution with a mean of 0 and an accuracy of α0 to Gam(a,b), where Gam(·) represents a Gamma distribution, then the fth column vector in the receiving antenna array data Y satisfies represents complex Gaussian distribution, I M represents the M-dimensional identity matrix;
[0028] Assume that the fth column vector x in the sparse signal matrix X is f The mean is 0 and the precision is k=1…,K,where K represents the number of categories and diag(·) represents vector matrixization. Then x f The probability density function of is:
[0029]
[0030] in, 1[z f =k] means when z f = k, the value is 1, otherwise the value is 0; f is the distribution vector, subject to the parameters {ω k} k=1,…,K The multinomial distribution of is the weight of the "stick-cutting" theorem, and the length of the stick-cutting corresponding to K times of stick-cutting is [π1,...,π K ],π k ~Beta(1,γ) is the length of the “cut stick” theorem; γ~Gam(e,h) is the scaling parameter of the Dirichlet process.
[0031] The present invention is further configured as follows: Step S3 comprises:
[0032] For each subband, a strong constraint virtual node h is introduced f =Ax f , f=1,…,F.
[0033] The present invention is further configured as follows: Step S4 comprises:
[0034] The sparse signal matrix X is updated by the GAMP algorithm, and the fth column vector of X is x f ,f=1,…,F, the mean of the Gaussian distribution it obeys is μ f and the precision matrix is Γ f :
[0035] According to the GAMP algorithm, update the intermediate variables and X in sequence:
[0036]
[0037] Where α0 is the noise accuracy, 1[z f =k] means when z f = k, the value is 1, otherwise the value is 0. Indicated by The diagonal matrix formed by <·> indicates expectation; (·) H represents the conjugate transpose.
[0038] The update formula of noise accuracy α0 is:
[0039]
[0040] parameter The update formula is:
[0041]
[0042] In the formula, for The nth value of f,n μ f The nth value of ; for The nth value of ;
[0043] The stick theorem variable π k The update formula is:
[0044]
[0045] In the formula,
[0046] The update formula of the scale parameter γ is:
[0047]
[0048] Where Ψ(·) is the Digamma function;
[0049] Assign variable z f The update formula is:
[0050]
[0051] Where tr(·) represents the trace of the matrix;
[0052] The present invention is further configured as follows: Step S5 comprises:
[0053] For the parameter z in the probability model f , γ, intermediate variable: p n ,v s ,s, q f , h f Assign an initial value, substitute the data Y of the receiving antenna array into the update formula of step S5, and continuously iterate and update until a preset threshold is reached or a preset maximum number of iterations is reached, and then stop updating to obtain the final parameter estimation value.
[0054] The present invention is further configured as follows: Step S6 comprises:
[0055] Calculate the fth column vector x of the sparse signal matrix X f Estimated value of Find The position of the maximum value of , the angle of the grid corresponding to this position is the DOA estimation value.
[0056] The present invention is further configured to introduce a strong constraint on the off-grid angle deviation for each grid: a(θ n )+b(θ n )(θ n -θ k ); is the incident angle θ of the true broadband signal k The real steering vector is k Nearest neighbor grid θ n At the point, the first-order Taylor expansion is used to obtain b(θ n ) is a(θ n ) is the derivative of .
[0057] The present invention has the following beneficial effects:
[0058] The present invention firstly establishes a sparse signal model of the received signal based on the characteristics of the broadband incident signal based on a uniform linear array, then constructs a probability model based on the Dirichlet process prior for each variable in the model, and then, by introducing virtual nodes, proposes a variational Bayesian broadband DOA estimation method based on the GAMP algorithm; after initializing the unknown parameters, input the receiving antenna array data Y, iterate and update iteratively through the DP-GAMP algorithm, and obtain the estimated values of the probability model parameters; finally, according to the obtained parameter estimation values, the estimated value of DOA is calculated. The method has the characteristics of low complexity and high estimation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] Figure 1 It is a flow chart of a 6G ISAC system broadband signal DOA estimation algorithm provided by the present invention;
[0060] Figure 2 It is a schematic diagram of the receiving antenna array of the present invention. DETAILED DESCRIPTION
[0061] The present invention is further described in detail below in conjunction with specific embodiments and drawings, but the embodiments of the present invention are not limited thereto.
[0062] The flowchart of a 6G ISAC system broadband signal DOA estimation algorithm provided by the present invention is as follows: Figure 1 As shown, the adopted technical solution is divided into the following 5 steps:
[0063] S1: Establish a sparse signal model for broadband signals received by a uniform linear array, as follows:
[0064] M sensors form a uniform linear array with a spacing of Δd, receive the broadband signal transmitted by K antennas in the far field, and divide the broadband signal into F narrowband signals through filters to form an M×F dimensional output signal matrix Y;
[0065] Then based on the angle domain, the range [0 ° ,180 ° ] is evenly divided into N grids, and the angle grid point set Θ = [θ1,…,θ N ], then define the broadband signal sent by K antennas as N×F dimensional sparse signal matrix X in the angle domain, define M×F dimensional noise signal matrix W, and define M×N dimensional overcomplete dictionary matrix A;
[0066] The sparse signal model of the output signal matrix Y is constructed based on the sparse signal matrix X, the noise signal matrix W and the overcomplete dictionary matrix A:
[0067] Y=AX+W
[0068] in,
[0069] A=[a(θ1),…,a(θ i ),...,a(θ N )],
[0070]
[0071] In the formula, a(θ i ), i=1,…,N means the incident angle is θ i is the guiding vector at , and λ is the wavelength.
[0072] In this embodiment, for each grid, a strong constraint on the off-grid angle deviation is introduced. is the incident angle θ of the true broadband signal k The real steering vector is k Nearest neighbor grid θ n At the point, the first-order Taylor expansion is used to obtain b(θ n ) is a(θ n ) is the derivative of .
[0073] Step 2: Establish a probabilistic broadband signal model based on the Dirichlet process prior. The details are as follows:
[0074] Assuming that the additive noise of the broadband signal obeys a complex Gaussian distribution with a mean of 0 and an accuracy of α0~Gam(a,b), where Gam(·) represents a Gamma distribution, then the fth column vector in the receiving antenna array data Y satisfies represents complex Gaussian distribution, I M represents the M-dimensional identity matrix;
[0075] Assume that the fth column vector x in the sparse signal matrix X is f The mean is 0 and the precision is k=1...,K,where K represents the number of categories and diag(·) represents vector matrixization. Then x f The probability density function of is:
[0076]
[0077] in, 1[z f =k] means when z f = k, the value is 1, otherwise the value is 0; f is the distribution vector, subject to the parameters {ω k} k=1,...,K The multinomial distribution of is the weight of the "stick-cutting" theorem, and the length of the stick-cutting corresponding to K times of stick-cutting is [π1,...,π K ]; π k ~Beta(1,γ) is the length of the “cut stick” theorem; γ~Gam(e,h) is the scaling parameter of the Dirichlet process.
[0078] Step 3: Introduce virtual variable nodes, as follows:
[0079] For each subband, a strong constraint virtual node h is introduced f =Ax f , f=1,…,F;
[0080] Step 4: Based on the variational Bayesian broadband DOA estimation method of the GAMP algorithm, the update formulas of each variable are obtained, as follows:
[0081] The sparse signal matrix X is updated by the GAMP algorithm, and the fth column vector of X is x f ,f=1,…,F, the mean of the Gaussian distribution it obeys is μ f and the precision matrix is Γ f :
[0082] According to the GAMP algorithm, update the intermediate variables and X in sequence:
[0083]
[0084] Where α0 is the noise accuracy, 1[z f =k] means when z f = k, the value is 1, otherwise the value is 0. Indicated by The diagonal matrix formed by <·> indicates expectation; (·) H represents the conjugate transpose.
[0085] The update formula of noise accuracy α0 is:
[0086]
[0087] parameter The update formula is:
[0088]
[0089] In the formula, for The nth value of f,n μ f The nth value of ; for The nth value of ;
[0090] The stick theorem variable π k The update formula is:
[0091]
[0092] In the formula,
[0093] The update formula of the scale parameter γ is:
[0094]
[0095] Where Ψ(·) is the Digamma function;
[0096] Assign variable z f The update formula is:
[0097]
[0098] Where tr(·) represents the trace of the matrix;
[0099] Step 5: Initialize the unknown parameters, input the received output signal matrix Y, and iteratively update through the DP-GAMP algorithm to obtain the estimated values of the probability model parameters, as follows:
[0100] For the parameter z in the probability model f , γ, intermediate variable: p n ,v s ,s, q f , h f , assign an initial value, substitute the data Y of the receiving antenna array into the update formula of step 5, and continuously iterate and update until the preset threshold is reached or the preset maximum number of iterations is reached, and then stop updating to obtain the final parameter estimation value;
[0101] Step 6: Based on the obtained parameter estimates, calculate the estimated value of DOA as follows:
[0102] Calculate the fth column vector x of the sparse signal matrix X f Estimated value of Find The position of the maximum value of , the angle of the grid corresponding to this position is the DOA estimation value.
[0103] Numerical simulation is performed using a computer to verify the estimation performance of the method proposed in the present invention.
[0104] The number of array elements of the uniform linear array receiving antenna used in the simulation is 10, and the array element spacing is half the wavelength of the incident signal. The angle range of [0°, 180°] is evenly divided into 181 grids at intervals of 1° to obtain an angle grid point set. In order to verify the effectiveness of the method proposed in this paper, DOA estimation is performed under different simulation conditions and the simulation experiment results are plotted.
[0105] The above are only preferred embodiments of the present invention. It should be noted that the above preferred embodiments should not be regarded as limiting the present invention. Any simple modification made to the above embodiments according to the technical essence of the present invention belongs to the scope of the present invention. The protection scope of the present invention shall be based on the scope defined by the claims.
Claims
1. A 6G ISAC system wideband signal DOA estimation algorithm, characterized in that: The following steps are involved: S1: Use M sensors to build a uniform linear array, receive the broadband signals transmitted by K far-field antennas, and divide the K broadband signals into F narrowband signals, where K is a positive integer, to form the output signal matrix Y; Based on the angle domain, the range [0°, 180°] is evenly divided into N grids to obtain the angle grid point set Θ = [θ1, ..., θ N ], and then define the broadband signal sent by K antennas as an N×F dimensional sparse signal matrix X in the angle domain, define an M×F dimensional noise signal matrix W, and define an M×N dimensional overcomplete dictionary matrix A; A sparse signal model of the output signal matrix Y is constructed based on the sparse signal matrix X, the noise signal matrix W and the overcomplete dictionary matrix A; S2: Based on the constructed sparse signal model of the output signal matrix Y, a probabilistic broadband signal model based on the Dirichlet process prior is established; S3: For each subband, introduce a virtual variable node h f =Ax f , f=1,…,F; S4: A variational Bayesian broadband DOA estimation method based on the GAMP algorithm is proposed, namely the DP-GAMP algorithm; The step S4 includes: updating the sparse signal matrix X by using the GAMP algorithm, where the fth column vector of X is x f ,f=1,...,F, the mean of the Gaussian distribution it obeys is μ f and the precision matrix is Γ f : According to the GAMP algorithm, update the intermediate variables and X in sequence: Where α0 is the noise accuracy, 1[z f =k] means when z f = k, the value is 1, otherwise the value is 0, f To allocate a vector, Indicated by The diagonal matrix formed by <·> indicates expectation; (·) H represents conjugate transpose; S5: Initialize the unknown parameters, input the received output signal matrix Y, iteratively update through the DP-GAMP algorithm, and obtain the estimated values of the probability model parameters; S6: Calculate the estimated value of DOA based on the obtained parameter estimated values.
2. A 6G ISAC system wideband signal DOA estimation algorithm according to claim 1, characterized in that: The step S1 comprises: M sensors form a uniform linear array with a spacing of Δd to receive broadband signals transmitted by K far-field antennas; The broadband signal is divided into F narrowband signals through a filter to form an M×F dimensional output signal matrix Y; Then, based on the angle domain, the range [0°, 180°] is evenly divided into N grids to obtain the angle grid point set Θ = [θ1,…,θ N ]; Then define the broadband signal sent by K antennas as N×F dimensional sparse signal matrix X in the angle domain, define M×F dimensional noise signal matrix W, and define M×N dimensional overcomplete dictionary matrix A; Based on the sparse signal matrix X, the noise signal matrix W and the overcomplete dictionary matrix A, the sparse signal model of the output signal matrix Y is constructed as follows: Y=AX+W in, A=[a(θ1),…,a(θ i ),...,a(θ N )], In the formula, a(θ i ), i=1,...,N means the incident angle is θ i is the guiding vector at , and λ is the wavelength.
3. A 6G ISAC system wideband signal DOA estimation algorithm according to claim 2, characterized in that: Probabilistic broadband signal models based on Dirichlet process priors include: Assuming that the additive noise of the broadband signal obeys a complex Gaussian distribution with a mean of 0 and an accuracy of α0 to Gam(a,b), a and b represent hyperparameters greater than 0, where Cam(·) represents a Gamma distribution, then the fth column vector in the receiving antenna array data Y satisfies represents complex Gaussian distribution, I M represents the M-dimensional identity matrix, assuming that the f-th column vector x in the sparse signal matrix X f The mean is 0 and the precision is The complex Gaussian distribution of x, where K represents the number of categories and diag(·) represents the vector matrix. f The probability density function of is: in, 1[z f =k] means when z f = k, the value is 1, otherwise the value is 0, f is the distribution vector, subject to the parameters {ω k } k=1,…,K The multinomial distribution of is the weight of the "stick-cutting" theorem, π k ~Beta(1,γ) is the length of the "stick" theorem; γ~Gam(e,h) is the scaling parameter of the Dirichlet process.
4. The 6G ISAC system wideband signal DOA estimation algorithm according to claim 1, characterized in that: The step S4 further comprises: The update formula of noise accuracy α0 is: parameter The update formula is: In the formula, for The nth value of f,n μ f The nth value of ; for The nth value of ; The stick theorem variable π k The update formula is: In the formula, The update formula of the scale parameter γ is: Among them, Ψ(·) is the Digamma function; Assign variable z f The update formula is: Where tr(·) represents the trace of the matrix.
5. The 6G ISAC system wideband signal DOA estimation algorithm according to claim 4, characterized in that: The step S5 comprises: For the parameter z in the probability model f , γ, intermediate variable: p n ,v s ,s, q f , h f Assign an initial value, substitute the data Y of the receiving antenna array into the update formula of step S5, and continuously iterate and update until a preset threshold is reached or a preset maximum number of iterations is reached, and then stop updating to obtain the final parameter estimation value.
6. The 6G ISAC system wideband signal DOA estimation algorithm according to claim 1, characterized in that: The step S6 comprises: Calculate the fth column vector x of the sparse signal matrix X f Estimated value of Find The position of the maximum value of , the angle of the grid corresponding to this position is the DOA estimation value.
7. The 6G ISAC system wideband signal DOA estimation algorithm according to claim 1, characterized in that: For each grid, a strong constraint on off-grid angle deviation is introduced is the incident angle θ of the true broadband signal k The real steering vector is k Nearest neighbor grid θ n At the point, the first-order Taylor expansion is used to obtain b(θ n ) is a(θ n ) is the derivative of .
