A robust symbol detection method in the presence of impulse noise

Abstract

The application discloses a kind of robust symbol detection methods under pulse noise environment, belong to data underwater acoustic communication technical field, based on vector approximate message passing and variational bayes inference symbol detection algorithm.The algorithm models the symbol detection model in block processing as the combination of normal signal and abnormal value of pulse interference signal, and uses a set of binary indicator variables to automatically identify abnormal values, under this combination model, all variables are estimated using variational bayes inference, and the pulse noise is compensated between the estimated communication signals, thereby improving the stability of the algorithm in the presence of pulse noise.The algorithm proposed by the application can obtain better performance than traditional algorithms without increasing additional computational complexity.The application can significantly reduce the influence of pulse noise interference on traditional symbol detection algorithms.

Description

Technical Field

[0001] This invention belongs to the field of underwater acoustic communication technology, specifically relating to a robust symbol detection method under impulse noise environment. Background Technology

[0002] Compared to multi-carrier systems, single-carrier systems do not suffer from a high peak-to-average power ratio (PAPR), and Doppler frequency offset is easier to estimate and compensate for. Therefore, they are one of the best solutions for realizing high-speed underwater acoustic (UWA) communication. However, single-carrier systems are still affected by adverse factors such as multipath delay and time-varying channels. To combat the harsh conditions of UWA channels, turbo equalization, as a powerful symbol detection technique, has received considerable attention in recent years. To further improve equalization performance under the same complexity, message-passing-based equalizer design has been extensively studied, including algorithms such as Approximate Message Passing (AMP), Generalized Approximate Message Passing (GAMP), Vector Approximate Message Passing (VAMP), Unitary Approximate Message Passing (UAMP), and Orthogonal Approximate Message Passing (OAMP).

[0003] However, the above methods all assume that the background noise has a Gaussian distribution, which is not the case in reality. With the increasing frequency of human maritime activities and the expansion of information transmission scenarios, UWA communication systems face performance degradation due to impulse noise. Generally, the probability density function of impulse noise has a heavier tail than Gaussian noise, which may cause bias or even failure in traditional algorithms. In the field of communication, impulse noise suppression has been extensively studied; however, relevant research mainly exists in Orthogonal Frequency Division Multiplexing (OFDM) systems, while research on single-carrier impulse noise suppression is scarce.

[0004] The most commonly used impulse noise cancellation method is the thresholding method, which can detect and eliminate sampling points exceeding a certain threshold. However, selecting a suitable nonlinear operating threshold to adequately suppress impulse noise without distorting the communication signal is often difficult. Besides the thresholding method, correlated iterative receivers have been developed, which can couple impulse noise; however, this requires some prior information about the impulse noise, which is usually difficult. Furthermore, leveraging the sparsity of impulse noise, correlated compressed sensing has been applied to impulse noise detection and cancellation, and this method has proven effective. However, when effective observation of impulse noise is lacking, compressed sensing algorithms can generate a large amount of noise in areas where there is no impulse noise, leading to inaccurate impulse noise compensation and thus distortion of the communication signal. Summary of the Invention

[0005] To address the technical problem of performance degradation in traditional symbol detection algorithms under impulse noise, this invention provides a robust symbol detection method for impulse noise environments. Based on Variational Bayesian Inference (VBI) and VAMP, this robust symbol detection algorithm models observations as a mixture of the target signal and outliers, using a set of binary index variables to identify whether an observation is an outlier. Then, under the proposed mixture model, VBI is used to alternately infer the posterior distributions of the target signal and impulse noise, compensating for the impulse noise when estimating the target signal. The VAMP algorithm is then embedded within the VBI framework, using VAMP to estimate the approximate symbol distribution in VBI, thereby improving the iterative performance of VBI. Conversely, VBI can also provide updated signal-to-noise ratio information for VAMP, resulting in additional gains compared to the traditional VAMP algorithm. This method achieves high-precision symbol detection under impulse noise environments with low computational complexity.

[0006] To solve the technical problem, the technical solution of the present invention is as follows:

[0007] A robust symbol detection method under impulse noise environment, the method comprising:

[0008] S1: Based on observation data Establish the received signal model and initialize the data block index i=1. I is the number of blocks, and the index of the variational Bayesian inference (VBI) iteration count is k=1. K is the maximum number of iterations in VBI, the index of the iteration number is iter=1, and ITER is the maximum number of iterations in turbo balancing.

[0009] S2: Prior knowledge of impulse noise Initialize the noise variance γ and the binary index variable z;

[0010] S3: Perform channel estimation using a pre-set training sequence x and the received observation data. Obtain the channel h and output the estimated channel matrix. ;

[0011] S4: Perform inter-block interference cancellation. Reconstruct the inter-block interference using data from the previous data block and the estimated channel, removing the inter-block interference from the observed data. Subtracting from the data in the middle yields the data containing the i-th data block and noise. ;

[0012] S5: Estimating communication signals ;

[0013] S6: Estimate impulse noise using the observed signal obtained after inter-block interference cancellation. Communication signals obtained from S5 This yields the estimated impulse noise result e; simultaneously, it updates the impulse noise prior. Related hyperparameters ,in Represents the expected value of a variable;

[0014] S7: Estimate the noise variance γ and the binary index variable z;

[0015] S8: Update , ;

[0016] S9: k=k+1, execute S5-S8 until the iteration converges;

[0017] S10: i = i + 1, execute S3-S9 until all data blocks have been checked;

[0018] S11: Demap the output of S5 to obtain the log-likelihood ratio information;

[0019] S12: Deinterleaving and channel decoding, decoder output bit sequence Information compared to likelihood , Indicates the decoded output bits;

[0020] S13: Interleaving becomes equalizer prior information ;

[0021] S14: Continue executing S1-S13 until the iteration count iter reaches the maximum iteration count ITER.

[0022] Furthermore, in step S3, a measurement matrix X is constructed using a preset training sequence x, and the measurement matrix X is then used in conjunction with the observed signal. Obtain channel h.

[0023] Furthermore, step S5 specifically includes:

[0024] S51: Denoising version of computational signal :

[0025] ;

[0026] S52: Calculation External information:

[0027]

[0028] S53: Calculate the Gaussian white noise version of the signal :

[0029]

[0030] Where I represents an identity matrix of appropriate size, and diag represents vector diagonalization;

[0031] S54: Calculation External information:

[0032] ;

[0033] The communication signal is obtained by iterating through the above operations. = .

[0034] Furthermore, in step S6, the impulse noise estimation specifically includes:

[0035] .

[0036] Furthermore, in step S7, the estimation of the noise variance γ specifically includes:

[0037] ;

[0038] in:

[0039] .

[0040] Furthermore, in step S7, the estimation of the index variable z specifically includes:

[0041] ;

[0042] in:

[0043] ;

[0044] Where C is the normalization constant.

[0045] Furthermore, in step S9, if k ≤ K, then steps S5-S8 are executed until convergence; if k > K, then the iteration stops and the estimated sign is output. And continue execution downwards.

[0046] Furthermore, in step S10, if i ≤ I, then steps S3-S9 are executed until all data blocks have been detected; if i > I, then the iteration stops and all data blocks are output. And continue execution downwards.

[0047] Furthermore, in step S11, the likelihood ratio information is calculated as follows:

[0048] ;

[0049] in, and This represents the output mean and variance of the equalizer. Indicates the modulation constellation point, Indicates the modulation order. This represents the log-likelihood ratio of the feedback from the previous iteration. This represents the encoded bit sequence.

[0050] Furthermore, in step S14, if iter ≤ ITER, then steps S1-S13 are executed until the maximum number of iterations is reached; if iter > ITER, then the iteration stops and the result is output. .

[0051] Compared with the prior art, the advantages of the present invention are as follows:

[0052] This invention automatically identifies the location of impulse noise by using binary indicator variables without requiring preprocessing of the impulse noise. By combining VAMP with VBI, it leverages the strong convergence performance of VAMP to estimate the posterior distribution of symbols, achieving high-precision symbol estimation with low complexity. Conversely, VBI provides updated signal-to-noise ratio information to VAMP, resulting in additional gain. Attached Figure Description

[0053] Figure 1 A flowchart of a single-carrier symbol detection algorithm based on vector approximation message passing and variational Bayesian inference under impulse noise environment;

[0054] Figure 2 Simulated channel diagram;

[0055] Figure 3 The actual time-domain waveform of the collected impulse noise;

[0056] Figure 4 (a) shows the bit error rate statistics under weak impulse interference, and (b) shows the bit error rate statistics under strong impulse interference. Detailed Implementation

[0057] The specific implementation of the present invention is described below with reference to embodiments:

[0058] It should be noted that the structures, proportions, sizes, etc. shown in this specification are only used to complement the content disclosed in the specification for those skilled in the art to understand and read, and are not intended to limit the conditions under which the present invention can be implemented. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0059] Furthermore, the terms such as "upper," "lower," "left," "right," "middle," and "one" used in this specification are merely for clarity of description and are not intended to limit the scope of the invention. Any changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention.

[0060] Example 1:

[0061] like Figure 1 As shown, the steps of the present invention are as follows:

[0062] (1) To establish a SISO single-carrier underwater acoustic communication system, the matrix form of the received signal model can be described as:

[0063]

[0064] Where i represents the block index. , Represents the length of the block. It is the i-th data block from the sending end. Represents a block of impulse noise. It is a Gaussian noise block. It is and channel The corresponding channel matrix can be represented as

[0065]

[0066] Where L is the length of the channel, and the index i=1 is initialized for the target data block. I is the number of blocks, and the VBI iteration index k=1. K is the maximum number of iterations in VBI, the index of the iteration number is iter=1, and the maximum number of iterations is ITER;

[0067] (2) Initialization: Initialize the impulse noise prior e, initialize the noise variance γ, and initialize the binary index variable z;

[0068] (3) Channel estimation: Construct a measurement matrix X using the known training sequence x, and then perform channel estimation based on the measurement matrix X and the observed signal. The channel h can be obtained, and the estimated channel matrix H can be output.

[0069] (4) Inter-block interference cancellation: Inter-block interference is reconstructed using data from the previous data block and the estimated channel, thus removing inter-block interference from the observed data. Subtracting from the middle yields data containing only the i-th data block and noise. ;

[0070] (5) Target symbol detection: using the channel matrix H in (3) and the observation data in (4) By combining impulse noise and indicator variables, the target symbol can be estimated. The specific operation of symbol estimation is as follows:

[0071] A. Denoising version of the computational signal

[0072]

[0073] B. Calculation External information

[0074]

[0075] C. Calculate the Gaussian white noise version of the signal.

[0076]

[0077] D. Calculation External information

[0078] ;

[0079] in The decoder feedback result from the previous iteration was uniformly distributed during the first iteration. By mapping constellation points to the transmitter and iterating through the above operations, the communication signal can be obtained. = ;

[0080] (6) Impulse noise estimation: The observed signal obtained after inter-block interference cancellation. The communication signal obtained in (5) The estimation results of impulse noise can be obtained:

[0081] ;

[0082] in, Simultaneously update the hyperparameters related to the impulse noise e. ,in Represents the expected value of a variable;

[0083] (7) Estimating the noise variance and the binary index variable z: The estimation result of the noise variance is as follows:

[0084]

[0085] in:

[0086]

[0087] The estimation results for the indicator variables are as follows:

[0088] ;

[0089] in:

[0090] ;

[0091] Where C is the normalization constant.

[0092] (8) Update : ;

[0093] (9) Judgment: If k≤K, then execute steps (5)-(8) until convergence; if k>K, then stop iterating and output the estimated sign. And execute downwards;

[0094] (10) Judgment: If i≤I, then execute steps (3)-(9) until all data blocks have been detected; if i>I, then stop the iteration and output all And continue executing downwards;

[0095] (11) Demapping: Demapping all By performing demapping, the log-likelihood ratio information is obtained. The likelihood ratio information is calculated as follows:

[0096] ;

[0097] in, and This represents the output mean and variance of the equalizer. Indicates the modulation constellation point, Indicates the modulation order. This represents the log-likelihood ratio of the feedback from the previous iteration. Represents the encoded bit sequence;

[0098] (12) Deinterleaving and channel decoding: decoder output bit sequence Information compared to likelihood , Indicates the decoded output bits;

[0099] (13) Interleaving becomes equalizer prior information ;

[0100] (14) Judgment: If iter≤ITER, then execute steps (1)-(13) until the maximum number of iterations is reached; if iter>ITER, then stop iterating and output. .

[0101] Example 2:

[0102] The simulation research of this invention specifically includes:

[0103] (1) Simulation conditions and parameters:

[0104] Table 1

[0105]

[0106] To simulate a realistic sparse underwater acoustic channel, we set most taps of the channel to 0 and the maximum multipath delay to 60ms. The channel is as follows: Figure 2 As shown in Table 1, the relevant parameter settings of the system are as follows. The impulse noise used in the simulation is the actual acquired impulse noise, and the time-domain waveform of the acquired impulse noise is shown in Table 1. Figure 3 As shown

[0107] (2) Simulation results display:

[0108] Table 2

[0109]

[0110] Table 2 shows the bit error rate performance of the three algorithms after adding impulse noise. VAMP is the traditional VAMP symbol detection algorithm, VBI-VAMP is the traditional VBI-based symbol detection algorithm, and Proposed-VAMP is the symbol detection algorithm proposed in this invention. Two different amplitudes of the collected impulse noise are added to the communication system, with the larger amplitude noise referred to as strong impulse interference and the smaller amplitude noise referred to as weak impulse interference. The simulation results are the average bit error rate of 200 different data blocks. It can be seen that the algorithm proposed in this invention achieves the best results under both strong and weak impulse interference.

[0111] Figure 4 The bit error rate (BER) performance under two impulse interference environments is further demonstrated. The bar chart represents the percentage of blocks falling within the specified BER range after each iteration, and the results for 200 data blocks are statistically analyzed. It is clear that the algorithm proposed in this invention has the best performance. Figure 4 (a) When impulse interference is weak, the traditional VAMP algorithm achieves a BER of 0 for nearly 50% of blocks after 4 iterations, but the performance improvement is limited with further iterations. In contrast, the method with impulse noise suppression shows better performance improvement with iterations, and achieves a BER of 0 for nearly 80% of blocks after 4 iterations. Specifically, the proposed method achieves a BER of 0 for 90% of blocks after 4 iterations. Figure 4(b) As impulse noise increases, the performance of traditional methods further declines, with less than 30% of blocks having a bit error rate (BER) less than 0.05, and the performance remains almost unchanged with increasing iterations. Conversely, methods with impulse noise suppression show significant performance improvement with iterations; VBI-VAMP achieves nearly 50% BER of 0 after 4 iterations. In contrast, the method proposed in this invention achieves better performance after 3 iterations, and after 4 iterations, the percentage of blocks with BER of 0 exceeds 60%. Clearly, the receiver proposed in this invention has stronger impulse noise cancellation capability and better symbol estimation capability.

[0112] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

[0113] Many other changes and modifications can be made without departing from the concept and scope of this invention. It should be understood that this invention is not limited to the specific embodiments, and the scope of this invention is defined by the appended claims.

Claims

1. A robust symbol detection method under impulse noise environment, characterized in that, The method includes: S1: Based on observation data Establish the received signal model and initialize the data block index i=1. I is the number of blocks, and the index of the variational Bayesian inference (VBI) iteration count is k=1. K is the maximum number of iterations in VBI, the index of the iteration number is iter=1, and ITER is the maximum number of iterations in turbo balancing. S2: Prior knowledge of impulse noise Initialize the noise variance γ and the binary index variable z; S3: Perform channel estimation using a pre-set training sequence x and the received observation data. Obtain the channel h and output the estimated channel matrix. ; S4: Perform inter-block interference cancellation. Reconstruct the inter-block interference using data from the previous data block and the estimated channel, removing the inter-block interference from the observed data. Subtracting from the middle yields the data containing the i-th data block and noise. ; S5: Estimating communication signals ; S6: Estimate impulse noise using the observed signal obtained after inter-block interference cancellation. Communication signals obtained from S5 This yields the estimated impulse noise result e; simultaneously, it updates the impulse noise prior. Related hyperparameters ,in Represents the expected value of a variable; S7: Estimate the noise variance γ and the binary index variable z; S8: Update , ; S9: k=k+1, execute S5-S8 until the iteration converges; S10: i = i + 1, execute S3-S9 until all data blocks have been checked; S11: Demap the output of S5 to obtain the log-likelihood ratio information; S12: Deinterleaving and channel decoding, decoder output bit sequence Information compared to likelihood , Indicates the decoded output bits; S13: Interleaving becomes equalizer prior information ; S14: Continue executing S1-S13 until the iteration count iter reaches the maximum iteration count ITER.

2. The robust symbol detection method under impulse noise environment according to claim 1, characterized in that, In step S3, a measurement matrix X is constructed using a preset training sequence x, and the measurement matrix X is then used in conjunction with the observation data. Obtain channel h.

3. The robust symbol detection method under impulse noise environment according to claim 1, characterized in that, In step S6, the impulse noise estimation specifically includes: 。 4. The robust symbol detection method under impulse noise environment according to claim 1, characterized in that, In step S7, the estimation of the noise variance γ specifically includes: ； in: 。 5. The robust symbol detection method under impulse noise environment according to claim 1, characterized in that, In step S7, the estimation of the indicator variable z specifically includes: ； in: ； Where C is the normalization constant.

6. The robust symbol detection method under impulse noise environment according to claim 1, characterized in that, In step S9, if k ≤ K, then execute steps S5-S8 until convergence; if k > K, then stop iterating and output the estimated sign. And continue execution downwards.

7. The robust symbol detection method under impulse noise environment according to claim 1, characterized in that, In step S10, if i ≤ I, then steps S3-S9 are executed until all data blocks have been detected; if i > I, then the iteration stops and all data blocks are output. And continue execution downwards.

8. The robust symbol detection method under impulse noise environment according to claim 1, characterized in that, In step S11, the likelihood ratio information is calculated as follows: ； in, and This represents the output mean and variance of the equalizer. Indicates the modulation constellation point, Indicates the modulation order. This represents the log-likelihood ratio of the feedback from the previous iteration. This represents the encoded bit sequence.

9. A robust symbol detection method under impulse noise environment according to claim 1, characterized in that, In step S14, if iter ≤ ITER, then execute steps S1-S13 until the maximum number of iterations is reached; if iter > ITER, then stop iterating and output the result. .

Images