An improved estimation method, system, device and medium for a block-sparse Bayesian-based MIMO-OTFS system

Abstract

An improved estimation method, system, device and medium for a MIMO-OTFS system based on block sparse Bayesian, the method of which converts a burst block sparse vector into a traditional block sparse vector by using a lifting transformation matrix; calculates the start and end positions of the burst block sparse vector according to the traditional block sparse vector, and finds the distribution of the non-zero blocks in the angle dimension by combining the angle burst block length, updates the angle dimension block length and the corresponding column of the block dictionary matrix; solves the burst block sparse channel vector by using the theory of block sparse Bayesian; the system, device and medium of the application perform channel estimation based on the improved estimation algorithm for the MIMO-OTFS system based on block sparse Bayesian; the application is applied to the wireless communication system under high-speed movement in the 5G standard, the vehicle networking system, and provides reliable data transmission basis for intelligent driving in various complex application scenarios in underwater acoustic communication technology; the application in the MIMO-OTFS system not only has higher channel estimation accuracy than the traditional algorithm, but also has significant advantages in calculation efficiency and adaptability, and is an effective method for solving high complexity channel estimation problems.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, and particularly relates to an improved estimation method, system, device and medium for a MIMO-OTFS system based on block sparse Bayesian. BACKGROUND

[0002] At present, China has the world's longest and fastest high-speed railway network, and high-speed rail has become an indispensable means of transportation in many people's daily travel. Therefore, how to provide diversified and high-quality wireless communication services in a high-speed mobile scenario has become one of the main challenges faced by future wireless communication systems. However, the channel in a high-speed mobile scenario has a time-frequency double selection characteristic, which brings more severe challenges to the accuracy and effectiveness of channel estimation.

[0003] In a high-speed mobile scenario, the time-frequency double selection characteristic of the channel will greatly affect the orthogonality of the subcarriers of the Orthogonal Frequency Division Multiplexing (OFDM) system, thereby causing a sharp decline in its performance. Therefore, the Orthogonal Time Frequency Space (OTFS) system, as an alternative modulation scheme of the OFDM system, is applied in a high-speed mobile scenario. Unlike the OFDM system, which modulates data symbols in the time-frequency domain, the OTFS modulates data symbol signals in a two-dimensional time-delay-Doppler channel independent of time. Since the time-delay-Doppler channel intuitively reflects the positions of each scatterer constituting the wireless communication link at this time, it significantly reduces the impact of the time-frequency double selection channel on the signal, thereby meeting the diversified and high-quality communication in a high-speed mobile environment.

[0004] In addition, in the development process of wireless communication systems, the rapid growth of capacity demand and the over-high consumption of spectrum resources have become increasingly serious. In order to further improve the spectrum utilization rate of the OTFS system, the Multiple Input Multiple Output (MIMO) technology is combined with the OTFS technology, and spatial multiplexing is used to improve the channel capacity and channel transmission rate in a high-speed mobile scenario. In the MIMO-OTFS system, the addition of multiple antennas introduces a spatial dimension into the time-delay-Doppler domain, which not only increases the complexity of the channel, but also greatly increases the difficulty of channel estimation.

[0005] Because the MIMO-OTFS system has good co-sparse characteristics in the delay-Doppler-angle domain and the burst block sparse characteristics in the angle domain, the channel estimation matrix can be converted into a compressed sensing sparse vector solving problem. The reconstruction algorithm based on the Bayesian framework occupies an important position in the compressed sensing reconstruction algorithm, so it can usually obtain more accurate sparse solutions than convex optimization algorithms and greedy algorithms. The commonly used algorithm is sparse Bayesian learning (SBL), and the SBL estimation algorithm does not fully utilize the block sparse characteristics.

[0006] The patent application with the publication number CN116471148A proposes a channel estimation method based on sparse Bayesian learning, which adopts two-layer sparse prior information, including angle domain sparse channel parameters and phase shift matrix. The description of the channel characteristics is relatively simple, and the sparse Bayesian learning (SBL) algorithm framework is adopted, but the block sparse characteristics are not fully utilized, and the algorithm performance is relatively low. SUMMARY

[0007] In order to overcome the defects of the prior art, the purpose of the present application is to provide an improved estimation method, system, device and medium for a MIMO-OTFS system based on block sparse Bayesian, which introduces a virtual angle dimension on the basis of signal reconstruction by block sparse Bayesian (BSBL) algorithm, calculates the angle dimension related vector and maps it to a high-dimensional block sparse vector by using the lifting transformation matrix, and then inversely quantizes it to a traditional block sparse vector, which can effectively process the sparsity problem of the angle dimension; combined with the edge optimization (BO) algorithm with low time cost of estimating hyper-prior parameters, to obtain faster convergence speed, reduce the time required for algorithm running, and thus reduce the complexity of the algorithm. The application of the present application in the MIMO-OTFS system not only outperforms the traditional algorithm in channel estimation accuracy, but also shows significant advantages in computational efficiency and adaptability, which is an effective method to solve the problem of high complexity channel estimation.

[0008] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions to realize it:

[0009] An improved estimation method for a MIMO-OTFS system based on block sparse Bayesian, characterized in that it specifically comprises the following steps:

[0010] Step S1. Utilize the lifting transformation matrix Convert the burst block sparse vector into a traditional block sparse vector D θi ;

[0011] Step S2. Calculate the start and end positions of the burst block sparse vector according to the traditional block sparse vector D θi , and find the distribution of the angle dimension non-zero block Λ combining the angle burst block length Dθi For the length d of the angular dimension block i Distribution of non-zero angular dimension blocks Λ θi And update the corresponding columns of the block dictionary matrix Φ;

[0012] Step S3. Based on steps S1 and S2, a hierarchical model of block sparse Bayes is used to model the channel vector, calculate the posterior probability density and likelihood function, and accurately solve the burst block sparse channel vector to achieve block sparse channel vector estimation in noisy environments.

[0013] The specific process of transforming the burst block sparse vector into a traditional block sparse vector using the lifting transformation matrix in step S1 is as follows:

[0014] Step 1.1) Calculate the angle dimension correlation vector in, y and y' are the block dictionary matrix and the received signal, respectively;

[0015] Step 1.2) The angle dimension correlation vector χ obtained in Step 1.1) i Using lifting transformation matrix Mapping bursty block sparse vectors to high-dimensional block sparse vectors improves the efficiency to N. t D×1 high-dimensional block sparse vector d θi =Ω H μ, which is then inversely quantized to obtain a traditional block sparse vector.

[0016] In step S2, the start and end positions of the burst block sparse vector are calculated based on the traditional block sparse vector, and the distribution of non-zero angle dimension blocks is found by combining the angle burst block length D. The specific process of updating the angle dimension block length, the distribution of non-zero angle dimension blocks, and the block dictionary matrix is ​​as follows:

[0017] Step 2.1) Obtain the traditional block sparse vector D from Step 1.2). θi Calculate the traditional block sparse vector D θi The relevant vector g is obtained by using the L2 norm of all row vectors. θi (r)=||D θi (r,:)||2, then through the relevant vector g θi The maximum element yields the start and end positions t of the sparse vector of the burst block. θi =argmax(g θi (r));

[0018] Step 2.2) Update the angle dimension correlation vector χ based on the angle burst block length D. i The corresponding angular dimension block length d iThen the start and end positions t of the burst block sparse vector obtained by step 2.1) are used to update the distribution Λ of the angle dimension non-zero block θi θi Finally, the block dictionary matrix Φ is updated, and the update formula is as follows:

[0019]

[0020] The specific process in the step S3 is as follows:

[0021] Step 3.1) The block dictionary matrix Φ obtained in step 2.2) is used to model the channel vector by using the hierarchical model of the block sparse Bayesian, and the sparse relationship is as follows:

[0022] y = Φh + ε

[0023] Wherein, y is a received signal; h is a block sparse signal, which is described by the block correlation parameter γ i and the block correlation matrix B i ; ε is a noise signal and obeys a complex Gaussian distribution with a mean of zero and a covariance of λ -1 I, λ = δ -2 is a noise precision parameter; γ i , B i and λ are called hyper-prior parameters;

[0024] Step 3.2) The posterior probability density P(x|y, γ i , B i , λ) of the block sparse signal h and the likelihood function P(y|γ i , B i , λ) are calculated, and the calculation formula of the mean μ, the covariance Σ and the variance C of the likelihood function is obtained:

[0025]

[0026] Wherein, P(x|y, γ i , B i , λ) is a complex Gaussian distribution with a mean of μ and a covariance of Σ, and P(y|γ i , B i , λ) is a complex Gaussian distribution with a mean of 0 and a covariance of C;

[0027] By calculation, we obtain:

[0028] μ = λΣΦ H y, Σ = (Γ -1 + λΦ H Φ) -1 , C = λ -1 I + ΦΓΦ H ​

[0029] In the above formula, the intra-block correlation diagonal matrix Γ=diag(γ1B1,γ2B2,…,γ g B g );

[0030] Step 3.3) For the prior parameters λ and B i Update accordingly; the cost function can be obtained from the likelihood function, and using the Woodbury matrix identity, we have:

[0031]

[0032] Find λ and B in the above equation. i The partial derivatives of λ and B are obtained. i Update formula:

[0033]

[0034] Step 3.4) For the prior parameter γ i Update: The edge optimization (BO) algorithm is used to calculate the hyperprior parameters. An alternative function is found as the lower bound of the cost function. Then, the extreme points of the alternative function are found to obtain the updated values ​​of the hyperprior parameters. The formula of the cost function is derived as follows:

[0035]

[0036] Where const represents irrelevant terms; for the above formula For γ i By taking the partial derivatives, we obtain the prior parameter γ. i The updated formula is:

[0037]

[0038] Step 3.5) For the intra-block correlation matrix B i Add a constraint: Use a single relevant parameter r i Symmetric positive definite matrix To the intra-block correlation matrix B i Perform a replacement; when the symmetric positive definite matrix When it is a Toeplitz matrix, it is considered a symmetric positive definite matrix. It can effectively reflect the intra-block correlation of the sparse signal h.

[0039]

[0040] In the above formula, the correlation parameter r of the block sparse signal h i Based on B obtained in step 3.3) i Calculations show that:

[0041]

[0042] wherein, is the average of the sub-diagonal elements of the intra-block correlation matrix B i , and is the average of the main diagonal elements of B i ; the average correlation parameter of the block-sparse signal h is derived from the correlation parameters r i of each block-sparse signal h, i.e.

[0043] Finally, the replaced intra-block correlation matrix of the block-sparse signal h, i.e. the symmetric positive definite matrix B

[0044] Step 3.6) Repeat the iteration to solve the equation by steps 3.1) to 3.5) for the burst block-sparse channel vector h new , if the algorithm ends and outputs the result of the burst block-sparse channel vector h new (Λ θi ) = μ, otherwise repeat the above steps 3.2) - 3.5).

[0045] An application of the improved estimation method of the above-mentioned block-sparse Bayesian-based MIMO-OTFS system is applied to the wireless communication system under high-speed movement in the 5G standard, applied to the vehicle networking system to provide reliable data transmission basis for intelligent driving, and applied to various complex application scenarios such as underwater acoustic communication technology.

[0046] A system of the improved estimation method of the above-mentioned block-sparse Bayesian-based MIMO-OTFS system, comprising:

[0047] A block-sparse vector conversion module is used in step S1 to convert the burst block-sparse vector into a traditional block-sparse vector D θi by using the lifting transformation matrix ;

[0048] A traditional block-sparse vector calculation module is used in step S2 to calculate the start and end positions of the burst block-sparse vector, and find the distribution of the angle-dimension non-zero blocks Λ θi by combining the angle-dimension block length D θi , the distribution of the angle-dimension non-zero blocks Λ θi , and the corresponding column of the block dictionary matrix Φ, and update the angle-dimension block length d i ;

[0049] A block sparse Bayesian solution module is used in step S3 to model the channel vector according to steps S1 and S2 by using a hierarchical model of block sparse Bayesian, to calculate the posterior probability density and the likelihood function, and to accurately solve the burst block sparse channel vector, thereby realizing the block sparse channel vector estimation in a noise environment.

[0050] An apparatus for an improved estimation method of a MIMO-OTFS system based on the above block sparse Bayesian, comprising:

[0051] A memory for storing a computer program;

[0052] A processor for implementing the improved estimation method of the MIMO-OTFS system based on the block sparse Bayesian when the computer program is executed.

[0053] A computer readable storage medium for storing a computer program, which can solve the high complexity channel estimation problem based on the improved estimation method of the MIMO-OTFS system based on the block sparse Bayesian when the computer program is executed by a processor.

[0054] Compared with the prior art, the advantages and positive effects of the present application are:

[0055] 1. The present application introduces a virtual angle dimension on the basis of signal reconstruction by the block sparse Bayesian (BSBL) algorithm, calculates the angle dimension correlation vector, maps it to a high-dimensional block sparse vector by using a lifting transformation matrix, and then inversely quantizes it into a traditional block sparse vector, so as to effectively process the sparsity problem of the angle dimension. Not only the ability to capture the sparsity of the angle dimension is enhanced, but also the start and end positions of the burst block sparse vector can be more accurately calculated, thereby further improving the accuracy of channel estimation and making the reconstruction probability of the signal higher.

[0056] 2. The present application combines the edge optimization (BO) algorithm with a lower time cost of estimating hyper-prior parameters to obtain a faster convergence speed, reduce the time required for algorithm running, and thereby reduce the complexity of the algorithm. By finding a substitute function as the lower bound of the cost function, and then finding the extreme value point of the substitute function to obtain the updated value of the hyper-prior parameter, the calculation complexity of the hyper-prior parameter is significantly reduced, and the algorithm can quickly converge, so that the entire algorithm is more efficient when processing large-scale channel estimation problems.

[0057] 3、Compared with other traditional algorithms, the algorithm is more suitable for high complexity channel estimation problem in MIMO-OTFS scene. When processing multi-dimensional and large-scale sparse channel estimation, the traditional algorithm often faces the problems of high computational complexity and slow convergence speed. The algorithm introduces virtual angle dimension and edge optimization algorithm, which not only improves the probability and accuracy of signal reconstruction, but also greatly reduces the computational complexity and convergence time, so as to show more superior performance in practical application.

[0058] In summary, the application of the application in the MIMO-OTFS system is not only superior to the traditional algorithm in channel estimation accuracy, but also shows significant advantages in computational efficiency and adaptability, which is an effective method to solve the problem of high complexity channel estimation. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed to be used in the embodiments of the application will be briefly introduced as follows. Obviously, the drawings described below are only 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 these drawings.

[0060] Figure 1 The method flowchart of the embodiment of the application.

[0061] Figure 2 The hierarchical model diagram of block sparse Bayesian of the embodiment of the application.

[0062] Figure 3 The relationship diagram between the NMSE performance of different estimation algorithms and SNR of the embodiment of the application.

[0063] Figure 4 The BER and SNR relationship diagram of different estimation algorithms under QPSK modulation of the embodiment of the application. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical scheme and advantages of the application more clear, the application will be further described in detail below combined with embodiments. It should be understood that the specific embodiments described here are only used to explain the application, and are not used to limit the application.

[0065] In view of the problems existing in the prior art, the application provides an improved estimation method of MIMO-OTFS system based on block sparse Bayesian, which will be described in detail below combined with the drawings.

[0066] In order to enable those skilled in the art to fully understand how the application is implemented, this part is an explanatory embodiment of the explanation and description of the technical scheme of the claims. In order to enable those skilled in the art to fully understand how the application is implemented, this part is an explanatory embodiment of the explanation and description of the technical scheme of the claims.

[0067] like Figure 1 As shown, the improved estimation algorithm based on the MIMO-OTFS system provided in this embodiment of the invention includes the following steps:

[0068] Step S1. Using the lifting transformation matrix Transform the burst block sparse vector into a traditional block sparse vector D. θi The specific process is as follows:

[0069] 1.1) Calculate the angle-dimensional correlation vector in y and y' are the dictionary matrix and the received signal, respectively;

[0070] 1.2) The angle dimension correlation vector χ obtained in 1.1) i By using the lifting transformation matrix Mapping bursty block sparse vectors to high-dimensional block sparse vectors improves the efficiency to N. t D×1 high-dimensional block sparse vector d θi =Ω H μ, which is then inversely quantized to obtain a traditional block sparse vector.

[0071] Step S2. Based on the traditional block sparse vector D θi Calculate the start and end positions of the sparse vector of the burst block, and find the distribution Λ of the non-zero angular dimension blocks by combining the angular burst block length D. θi For the length d of the angular dimension block i Distribution of non-zero angular dimension blocks Λ θi And the corresponding columns of the block dictionary matrix Φ are updated; the specific process is as follows:

[0072] 2.1) The traditional block sparse vector D obtained through 1.2) θi Calculate D θi The relevant vector g is obtained by using the L2 norm of all row vectors. θi (r)=||D θi (r,:)||2, then through g θi The maximum element yields the start and end positions t of the sparse vector of the burst block. θi =argmax(g θi (r));

[0073] 2.2) Update the angle dimension correlation vector χ based on the angle burst block length D. i The corresponding angular dimension block length d i Then, the start and end positions t of the burst block sparse vector obtained in step 2.1) are determined. θi To update the distribution Λ of non-zero angular dimension blocks θiFinally, the block dictionary matrix Φ is updated, and the update formula is as follows:

[0074]

[0075] Step S3. According to step S1 and step S2, a block sparse Bayesian hierarchical model is used to model the channel vector, the posterior probability density and the likelihood function are calculated, and the burst block sparse channel vector is accurately solved, so as to realize the estimation of the block sparse channel vector in a noise environment; the specific process is as follows:

[0076] 3.1) According to the dictionary matrix Φ obtained according to 2.2), a block sparse Bayesian hierarchical model is used to model the channel vector, as shown in Figure 2 , the sparse relationship is:

[0077] y = Φh + ε

[0078] Wherein, y is the received signal; h is the block sparse signal, which is described by the block correlation parameter γ i and the block correlation matrix B i ; ε is a noise signal and obeys a complex Gaussian distribution with a mean of zero and a covariance of λ -1 I, λ = δ -2 is a noise precision parameter; γ i , B i and λ are called hyper-prior parameters;

[0079] 3.2) By calculating the posterior probability density P(x|y, γ i , B i , λ) and the likelihood function P(y|γ i , B i , λ) of the block sparse signal h, the calculation formula of the mean μ, the covariance Σ and the variance C of the likelihood function is obtained.

[0080]

[0081] Wherein, P(x|y, γ i , B i , λ) is a complex Gaussian distribution with a mean of μ and a covariance of Σ, and P(y|γ i , B i , λ) is a complex Gaussian distribution with a mean of 0 and a covariance of C.

[0082] Through calculation, we can get:

[0083] μ = λΣΦ H y, Σ = (Γ -1 + λΦ H Φ) -1 , C = λ -1 I + ΦΓΦH

[0084] In the above formula, the block-related diagonal matrix Γ = diag(γ1B1, γ2B2, …, γpBp) is introduced. g B g ).

[0085] 3.3) Update the hyper-prior parameters λ and B i ; according to the likelihood function, the cost function can be obtained, and by using the Woodbury matrix identity, we have:

[0086]

[0087] The partial derivatives of the above formula with respect to λ and B i can be obtained, and the update formula of λ and B i is as follows:

[0088]

[0089] 3.4) Update the hyper-prior parameter γ i ; the BO (Bounded Optimization) algorithm is used to calculate the hyper-prior parameter, and a surrogate function is used as the lower bound of the cost function, and then the extreme point of the surrogate function is found to obtain the updated value of the hyper-prior parameter. After derivation, the formula of the cost function is as follows:

[0090]

[0091] where const is an irrelevant term; the partial derivative of the above formula with respect to γ i can be obtained, and the update formula of the hyper-prior parameter γ i is as follows:

[0092]

[0093] 3.5) Add constraints to the block-related matrix B i ; a symmetric positive definite matrix i with only correlation parameter r i is used to replace the block-related matrix B i ; when the symmetric positive definite matrix is a Toeplitz matrix, it is considered that the symmetric positive definite matrix can well reflect the block-relatedness of the block-sparse signal h.

[0094]

[0095] In the above formula, the correlation parameter r i of the block-sparse signal h can be obtained by calculation according to the B i obtained in 3.3):

[0096]

[0097] wherein, is the average of the sub-diagonal elements of the intra-block correlation matrix B i , and is the average of the main diagonal elements of B i ; the average correlation parameter of the block sparse signal h is obtained by averaging the correlation parameters r i of each block sparse signal h, that is:

[0098] Finally, the replaced intra-block correlation matrix of the block sparse signal h, that is, the symmetric positive definite matrix B

[0099] 3.6) Repeat iteration, and solve: the burst block sparse channel vector h new is solved by steps 3.1) to 3.5), if the algorithm ends and the result is output as the burst block sparse channel vector h new (Λ θi )=μ, otherwise repeat the above steps 3.2)-3.5).

[0100] In order to prove the creativity and technical value of the technical scheme of the present application, the following is the application embodiment and verification of the technical scheme of the claim on the specific product or related technology.

[0101] The present application can be used in the wireless communication system under high-speed movement in the 5G standard, such as ensuring reliable wireless communication service under high-speed movement environment of high-speed rail, applied to vehicle networking system to provide reliable data transmission basis for intelligent driving, and applied to underwater acoustic communication technology and many other complex application scenarios.

[0102] It should be noted that embodiments of the present application can be implemented by hardware, software, or a combination of software and hardware. The hardware portion can be implemented with special logic; the software portion can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a specially designed hardware. Those skilled in the art can understand that the above-mentioned devices and methods can be implemented using computer executable instructions and / or included in processor control code, such as provided on a carrier medium, such as a magnetic disk, CD or DVD-ROM, programmable memory, such as read-only memory (firmware), or data carrier, such as optical or electronic signal carrier. The device of the present application and its modules can be implemented by hardware circuit, such as ultra-large scale integrated circuit or gate array, semiconductor, such as logic chip, transistor, etc., or programmable hardware device, such as field programmable gate array, programmable logic device, etc., or by software executed by various types of processors, or by a combination of the above-mentioned hardware circuit and software, such as firmware.

[0103] During the development or use of the embodiments of the present application, some positive effects have been achieved, and compared with the prior art, the present application indeed has great advantages, which will be described below in combination with the data and graphs of the test process.

[0104] 1. Simulation conditions

[0105] The simulation platform is built in MATLAB software for simulation test. The channel model adopts the standard space channel model of 3GPP, i.e. the urban macro unit environment of SCM channel model, and the specific parameter configuration is shown in the following table:

[0106] Parameter Parameter value Carrier frequency (GHz) 2.15 Subcarrier spacing (kHz) 15 OTFS frame size M x N 256×64 Channel model 3GPP SCM Array antenna ULA User speed (m / s) 100 Number of base station side antennas 8 Number of user antennas 1 Data modulation QPSK

[0107] 2. Simulation content and results

[0108] The performance simulation mainly compares the performances of the three algorithms:

[0109] 1) The existing BOMP estimation algorithm.

[0110] 2) The existing SBL estimation algorithm.

[0111] 3) The estimation algorithm of the present application (improved estimation algorithm of MIMO-OTFS system based on block sparse Bayesian, MABSBL).

[0112] Figure 3 The normalized mean square error (NMSE) performance comparison of the BOMP estimation algorithm, the SBL estimation algorithm and the proposed MABSBL estimation algorithm under different signal-to-noise ratios is shown. Figure 3It can be seen that the MABSBL algorithm exhibits excellent performance at all SNRs, with its NMSE curve lying at the lowest position, indicating that the MABSBL algorithm has the smallest error in the whole SNR range. In contrast, the SBL algorithm outperforms the BOMP algorithm, with its NMSE curve lying between those of the BOMP and MABSBL algorithms. The BOMP algorithm performs relatively poorly at all SNRs, with its NMSE curve lying at the highest position in the figure, reflecting that it has the largest signal estimation error. At an SNR of 0 dB, the NMSE of the MABSBL algorithm is about 10~1.5, while those of the BOMP and SBL algorithms are 10~0.7 and 10~1, respectively, showing the superiority of the MABSBL algorithm at low SNRs. As the SNR increases, the NMSEs of the three algorithms all decrease, but the MABSBL algorithm still performs the best. For example, at an SNR of 20 dB, the NMSE of the MABSBL algorithm decreases to 10~3, while those of the BOMP and SBL algorithms decrease to 10~2.1 and 10~2.7, respectively. This result shows that the MABSBL algorithm outperforms the other two algorithms in terms of NMSE performance at high SNRs.

[0113] Figure 4The BER performance of the MMSE detection algorithm for the signal is compared by using the channel state information obtained by the BOMP estimation algorithm, the SBL estimation algorithm and the MABSBL estimation algorithm under QPSK modulation and the channel state information under ideal conditions. It can be obviously seen from the figure that the MABSBL estimation algorithm exhibits the lowest BER under all signal-to-noise ratio conditions. Under the condition of a signal-to-noise ratio of 5 dB, the BER of the MABSBL estimation algorithm is about 10^-1.9, which is slightly better than 10^-1.8 of the BOMP estimation algorithm and 10^-1.9 of the SBL estimation algorithm. With the increase of the signal-to-noise ratio, the BER of the three algorithms presents a downward trend, but the decline speed and amplitude of the MABSBL estimation algorithm are more significant. Under a high signal-to-noise ratio of 20 dB, the BER of the MABSBL estimation algorithm decreases to about 10^-4.9, which is significantly better than 10^-3.9 of the BOMP estimation algorithm and 10^-4.6 of the SBL estimation algorithm, and even reaches a performance improvement of one order of magnitude compared with the BOMP, and is closer to the BER performance under the ideal state. This result shows that in the high signal-to-noise ratio region, the performance advantage of the MABSBL estimation algorithm is more obvious, and compared with the low signal-to-noise ratio region, the difference between the BER performance curve under the ideal state is smaller. In contrast, the BER performance of the BOMP estimation algorithm is relatively poor in the high signal-to-noise ratio region due to the poor NMSE performance in the low signal-to-noise ratio region, and a large gap is generated with the BER performance curve under the ideal state. In general, the MABSBL estimation algorithm exhibits better BER performance in the whole signal-to-noise ratio range, especially in the high signal-to-noise ratio region, which is closer to the performance under the ideal state.

[0114] In summary, the improved estimation algorithm of the MIMO-OTFS system based on block sparse Bayesian of the application has a great performance improvement compared with the existing BOMP estimation algorithm and BML estimation algorithm.

[0115] The above is only a specific embodiment of the application, but the protection scope of the application is not limited thereto, and any modification, equivalent replacement and improvement made by any person skilled in the art within the technical range disclosed by the application and within the spirit and principles of the application shall be covered within the protection scope of the application.

Claims

1. An improved estimation method for block-sparse Bayesian based MIMO-OTFS system, characterized in that: Specifically comprising the following steps: Step S1. Utilizing the lifting transform matrix Converting the burst block sparse vector into a conventional block sparse vector D θi ; Step 1.1) Compute angle-dimension dependent vectors where, and y are the block dictionary matrix and the received signal, respectively; Step 1.2) The angle dimension related vector χ obtained in step 1.1) is mapped to a high dimension block sparse vector d i using the lifting transformation matrix The burst block sparse vector is mapped to a high dimension block sparse vector, lifting to N t Dxl high dimension block sparse vector d θi = Ω H μ, and then inverse quantized to get the conventional block sparse vector Step S2. Compute the start and end positions of the burst block sparse vector D θi Compute the start and end positions of the burst block sparse vector D θi Update the distribution of angle-dimension non-zero blocks Λ i Update the distribution of angle-dimension non-zero blocks Λ θi and the corresponding columns of the block dictionary matrix Φ. Step 2.1) Obtain the traditional block sparse vector D from Step 1.2). θi Calculate the traditional block sparse vector D θi The relevant vector g is obtained by using the L2 norm of all row vectors. θi (r)=||D θi (r,:)||2, then through the relevant vector g θi The maximum element yields the start and end positions t of the sparse vector of the burst block. θi =argmax(g θi (r)); Step 2.2) Update the angle-dimension related vector χ according to the angle burst block length D i The corresponding angle-dimension block length d i Then update the distribution of angle-dimension non-zero blocks Λ by the start and end positions t of the burst block sparse vector obtained by step 2.1) θi θi Finally, update the block dictionary matrix Φ, and the update formula is as follows:​ Step S3. According to step S1, step S2, the block sparse Bayesian hierarchical model is used to model the channel vector, the posterior probability density and the likelihood function are calculated, and the burst block sparse channel vector is accurately solved, so as to realize the block sparse channel vector estimation in a noise environment.

2. The improved estimation method of a block-sparse Bayesian based MIMO-OTFS system according to claim 1, characterized in that: The specific process of the step S3 is as follows: Step 3.1) According to the block dictionary matrix Φ obtained in step 2.2), the block sparse Bayesian hierarchical model is used to model the channel vector, and the sparse relationship is: y=Φh+ε where y is the received signal; h is the block-sparse signal with intra-block correlation parameters γ i and the intra-block correlation matrix B i are collectively described; ε is the noise signal and follows a complex Gaussian distribution with mean zero and covariance λ -1 I, λ = δ -2 is the noise precision parameter; γ i , B i and λ are called hyper-prior parameters; Step 3.2) The mean μ, the covariance Σ and the variance C of the likelihood function are computed by calculating the posterior probability density P(x|y, γ i ,B i ,λ) and the likelihood function P(y|γ i ,B i ,λ) of the block-sparse signal h. Where, P(x|y,γ) i B i Let P(y|γ) be a complex Gaussian distribution with mean μ and covariance Σ. i B i ,λ) is a complex Gaussian distribution with mean 0 and covariance C; Through calculation, we get: μ = λ∑Φ H y,∑ = (Γ -1 + λΦ H Φ) -1 , C = λ -1 I + ΦΓΦ H In the above equation, the block-wise correlation diagonal matrix Γ = diag(γ1B1, γ2B2,..., γpBp) is given by g B g ); Step 3.3) Update the hyper-parameters λ and B i The cost function can be derived from the likelihood function, and using the Woodbury matrix identity, we have: Taking the partial derivatives of the above equation with respect to λ and B gives the update equations for λ and B i i ​​ Step 3.4) Update the hyperparameter γ i The BO algorithm is used to calculate the hyperparameter, and a surrogate function is used as the lower bound of the cost function. Then the extreme point of the surrogate function is found to obtain the updated value of the hyperparameter. The formula of the cost function is derived as follows: where const is the irrelevant term; and For γ i The partial derivative is obtained, and the updating formula of the hyperparameter γ i is: Step 3.5) Replace the intra-block correlation matrix B by a symmetric positive definite matrix i with a correlation parameter r only i Step 4) Replace the intra-block correlation matrix B by a symmetric positive definite matrix with a correlation parameter r only i Step 5) Replace the intra-block correlation matrix B by a symmetric positive definite matrix with a correlation parameter r only When the symmetric positive definite matrix is a Toeplitz matrix, it is considered that the symmetric positive definite matrix can well reflect the intra-block correlation of the block sparse signal h. In the above formula, the correlation parameter r of the block sparse signal h i B obtained according to step 3.3) i By calculation: wherein is the average of the off-diagonal elements of the block correlation matrix B i , and is the average of the main diagonal elements of B i ; the average correlation parameter of the block sparse signals h is derived from the correlation parameters r i of each block sparse signal h, i.e. Finally, the replaced block correlation matrix of the block sparse signal h, i.e., the symmetric positive definite matrix Step 3.6) Repeat iteration, solve: h = h + h by steps 3.1) to 3.5) on the burst block sparse channel vector h new Solve, if End of algorithm and output result: h new (Λ θi ) = μ, otherwise repeat above steps 3.2) - 3.5).

3. Use of the improved estimation method of the block-sparse Bayesian based MIMO-OTFS system according to any one of claims 1 to 2, characterized in that, It is applied to the wireless communication system under high-speed movement in the 5G standard, applied to the vehicle networking system to provide reliable data transmission basis for intelligent driving, and applied to the underwater acoustic communication technology and other complex application scenarios.

4. System for improved estimation method of a block-sparse Bayesian based MIMO-OTFS system according to any of claims 1 to 2, characterized in that It comprises: a block-sparse vector conversion module for converting the burst-sparse vector into a conventional block-sparse vector D in step S1 by using a lifting transformation matrix transforming the burst block-sparse vector into a conventional block-sparse vector D θi ; Step 1.1) Compute angle-dimension dependent vectors where, and y are the block dictionary matrix and the received signal, respectively; Step 1.2) Inverse quantization of the angle dimension related vector χ i , using the lifting transform matrix Mapping the burst block sparse vector into a high dimension block sparse vector, lifting to N t Dxl high dimension block sparse vector d θi = Ω H μ i , followed by inverse vectorization to get the conventional block sparse vector A traditional block-sparse vector computation module is used in step S2 to find the distribution of non-zero blocks in the angle dimension Λ by traditional block-sparse vector D θi The start and end positions of the burst block-sparse vector are calculated, and the distribution of non-zero blocks in the angle dimension Λ is found in combination with the angle dimension burst block length D θi The angle dimension block length d i The distribution of non-zero blocks in the angle dimension Λ θi And the corresponding column of the block dictionary matrix Φ is updated; Step 2.1) Obtain the traditional block sparse vector D from Step 1.2). θi Calculate the traditional block sparse vector D θi The relevant vector g is obtained by using the L2 norm of all row vectors. θi (r)=||D θi (r,:)||2, then through the relevant vector g θi The maximum element yields the start and end positions t of the sparse vector of the burst block. θi =arg max(g θi (r)); Step 2.2) Update the angle-dimension related vector χ according to the angle burst block length D i The corresponding angle-dimension block length d i Then update the distribution of angle-dimension non-zero blocks Λ by the start and end positions t of the burst block sparse vector obtained by step 2.1) θi θi Finally, update the block dictionary matrix Φ, the update formula is as follows:​ A block sparse Bayesian solving module, which is used in step S3 to model the channel vector according to step S1 and step S2 by using the block sparse Bayesian hierarchical model, calculate the posterior probability density and the likelihood function, accurately solve the burst block sparse channel vector, and realize the block sparse channel vector estimation in a noise environment.

5. Apparatus for improved estimation method of MIMO-OTFS system based on block sparse Bayesian of any of claims 1 to 2, characterized in that, It comprises: A memory for storing a computer program; A processor for executing the computer program to realize the improved estimation method of the MIMO-OTFS system based on block sparse Bayesian described in steps S1 to S3.

6. A computer-readable storage medium storing a computer program, the computer-readable storage medium being characterized by, When the computer program is executed by the processor, it can solve the high complexity channel estimation problem based on the improved estimation method of the MIMO-OTFS system based on block sparse Bayesian according to any one of claims 1 to 2.

Images