An adaptive interference suppression method under multi-load parallel communication

Abstract

The application relates to the technical field of wireless communication signal processing, in particular to an adaptive interference suppression method under multi-load parallel communication, which comprises the following steps: obtaining a baseband sampling data stream, performing frame boundary detection and synchronous head extraction to generate a synchronous multi-load symbol sequence; calling an improved minimum mean square error algorithm, fusing a time-varying channel autocorrelation matrix recursive updating rule of a recursive least square algorithm, estimating an interference covariance matrix, and avoiding a high complexity problem of traditional high-dimensional matrix inversion; according to the interference covariance matrix, constructing a space-time joint filter tap coefficient matrix, filtering the symbol sequence to obtain a parallel symbol stream after interference suppression; calculating a residual interference energy value and feeding back to the recursive updating step of the improved algorithm to correct the interference covariance matrix estimation value at the next moment. The method is suitable for fast time-varying channels, can reduce the calculation complexity while improving the dynamic adaptability of interference suppression, and is suitable for multi-load parallel communication scenes such as satellite communication and multi-beam networking.

Description

Technical Field

[0001] This invention relates to the field of wireless communication signal processing technology, and in particular to an adaptive interference suppression method for multi-load parallel communication. Background Technology

[0002] In multi-load parallel communication systems, intermodulation interference and co-channel interference are prone to occur when multiple signals share spectrum resources. Traditional solutions often employ static interference covariance matrix estimation combined with fixed-coefficient space-time filtering. Existing interference suppression methods based on minimum mean square error (MMSE) rely on matrix inversion operations, and the computational complexity increases exponentially in high-dimensional multi-load scenarios, making it difficult to meet real-time requirements. Furthermore, the time-varying characteristics of the channel cause a mismatch between the statically estimated interference matrix and the actual situation, resulting in a rapid degradation of filtering performance as the channel changes. In addition, conventional adaptive filtering systems lack a back-feedback mechanism, and the residual energy after interference suppression cannot be used to optimize the front-end estimation parameters, limiting the system's robustness to the accuracy of the initial estimation.

[0003] To address the aforementioned issues, it is necessary to solve the problem of efficient recursive updating of the interference covariance matrix under time-varying channels, as well as the problem of dynamic closed-loop optimization of interference suppression effectiveness. Although the traditional recursive least squares (RLS) algorithm has a fast convergence speed, when directly applied to multi-load scenarios, it still suffers from computational redundancy and insufficient stability due to the lack of coordinated design with the space-time filtering structure. On the other hand, the open-loop interference suppression system cannot adaptively adjust parameters according to the output quality, and is prone to performance fluctuations in complex electromagnetic environments. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an adaptive interference suppression method for multi-load parallel communication.

[0005] To achieve the above objectives, the present invention employs the following technical solution: an adaptive interference suppression method for multi-load parallel communication, comprising:

[0006] Acquire the baseband sampled data stream in a multi-payload parallel communication system, perform data frame boundary detection and synchronization header extraction on the baseband sampled data stream, and generate a synchronized multi-payload symbol sequence;

[0007] An improved least mean square error algorithm is invoked to perform interference matrix estimation on the multi-load symbol sequence. The improved least mean square error algorithm is based on the recursive least squares algorithm and generates an estimated value of the interference covariance matrix according to the autocorrelation matrix update rule of the time-varying channel.

[0008] Based on the estimated value of the interference covariance matrix, a tap coefficient matrix of a space-time joint filter is constructed. The multi-load symbol sequence is then filtered by the space-time joint filter to obtain a parallel symbol stream with interference suppression.

[0009] The residual interference energy is calculated on the parallel symbol stream after interference suppression to generate a residual interference energy value. The residual interference energy value is then fed back to the recursive update step of the improved minimum mean square error algorithm to correct the estimated value of the interference covariance matrix at the next time step.

[0010] As a further aspect of the present invention, the step of acquiring the baseband sampling data stream in a multi-payload parallel communication system, performing data frame boundary detection and synchronization header extraction processing on the baseband sampling data stream, and generating a synchronized multi-payload symbol sequence includes:

[0011] The baseband sampled data stream is subjected to matched filtering to generate a matched-filtered baseband complex envelope signal;

[0012] The matched-filtered baseband complex envelope signal is subjected to sliding cross-correlation operation with the locally stored frame synchronization sequence to generate a cross-correlation peak sequence. The peak position in the cross-correlation peak sequence that exceeds the detection threshold is used as the frame start position.

[0013] Based on the frame start position, a complete data frame is extracted from the baseband sampled data stream, and a synchronization header symbol segment is extracted from a specified position of the complete data frame;

[0014] The synchronization header symbol segment is compared with a preset synchronization header template to perform phase error estimation processing, generating carrier phase offset estimation value and timing error estimation value;

[0015] Based on the estimated carrier phase offset and the estimated timing error, the data symbol segments in the complete data frame are subjected to phase rotation and timing resampling to generate the synchronized multi-payload symbol sequence.

[0016] As a further aspect of the present invention, the synchronization header symbol segment is subjected to phase error estimation processing with a preset synchronization header template to generate carrier phase offset estimation value and timing error estimation value, including:

[0017] The complex value of each synchronization header symbol is read sequentially from the synchronization header symbol segment, and the standard complex symbol value at the corresponding position is read from the synchronization header template;

[0018] The complex value of each synchronization header symbol is multiplied by the conjugate of the corresponding standard complex symbol value to obtain the complex phase difference of each symbol position;

[0019] Perform an arctangent operation on the complex phase difference to obtain the instantaneous phase error value at each symbol position;

[0020] The instantaneous phase error values ​​of all symbol positions are arithmetically averaged to obtain the average phase error value, which is then used as the estimated carrier phase offset value.

[0021] The timing error difference between adjacent symbols is obtained by performing a differential operation between the phase difference between two adjacent symbols in the synchronization header symbol segment and the phase difference between two adjacent standard symbols in the synchronization header template. The sum of all timing error difference values ​​is used as the estimated timing error value.

[0022] As a further aspect of the present invention, the step of invoking the improved minimum mean square error algorithm to perform interference matrix estimation on the multi-load symbol sequence, wherein the improved minimum mean square error algorithm is based on the recursive least squares algorithm and generates an estimated value of the interference covariance matrix according to the autocorrelation matrix recursive update rule of the time-varying channel, including:

[0023] Initialize the forgetting factor, the inverse correlation matrix, and the weight vector of the recursive least squares algorithm;

[0024] The multi-load symbol sequence is sequentially input into the iterative update step of the recursive least squares algorithm in chronological order. The iterative update step includes: calculating the gain vector based on the input symbol vector at the current time and the inverse correlation matrix at the previous time; calculating the updated value of the weight vector at the current time based on the weight vector at the previous time, the gain vector, and the expected symbol value at the current time; and calculating the updated value of the inverse correlation matrix at the current time based on the inverse correlation matrix at the previous time and the gain vector.

[0025] After each iterative update step is completed, an interference coefficient matrix is ​​constructed based on the weight vector update value at the current time. The off-diagonal elements in the interference coefficient matrix are extracted, and all off-diagonal elements are arranged according to the load index to form the interference coefficient matrix at the current time.

[0026] Perform matrix multiplication on the current disturbance coefficient matrix and its conjugate transpose matrix to generate the current instantaneous disturbance covariance matrix.

[0027] Based on the forgetting factor, an exponentially weighted moving average is applied to the instantaneous disturbance covariance matrix at historical times and the instantaneous disturbance covariance matrix at the current time to generate an estimated value of the disturbance covariance matrix.

[0028] As a further aspect of the present invention, an exponentially weighted moving average is applied to the instantaneous disturbance covariance matrix at historical times and the instantaneous disturbance covariance matrix at the current time based on the forgetting factor to generate an estimated value of the disturbance covariance matrix, including:

[0029] After each iteration update step of the recursive least squares algorithm is completed, the instantaneous disturbance covariance matrix at the current moment is obtained;

[0030] Obtain the estimated value of the disturbance covariance matrix generated and stored in the previous iteration update step, and use it as the disturbance covariance matrix at the historical moment.

[0031] Read the currently valid forgetting factor value from the parameters of the recursive least squares algorithm, and calculate the first weighting coefficient and the second weighting coefficient, wherein the first weighting coefficient is equal to the forgetting factor value, and the second weighting coefficient is equal to one minus the forgetting factor value;

[0032] The interference covariance matrix at the historical moment is multiplied by the first weighting coefficient, and the instantaneous interference covariance matrix at the current moment is multiplied by the second weighting coefficient. The results of the two multiplication operations are then added to obtain the estimated value of the interference covariance matrix at the current moment.

[0033] The estimated value of the disturbance covariance matrix at the current moment is stored and used as the disturbance covariance matrix at the historical moment in the next iteration update step.

[0034] As a further aspect of the present invention, a tap coefficient matrix of a space-time joint filter is constructed based on the estimated value of the interference covariance matrix, and the multi-load symbol sequence is filtered by the space-time joint filter to obtain a parallel symbol stream with interference suppression, including:

[0035] The estimated value of the interference covariance matrix is ​​subjected to eigenvalue decomposition, and the eigenvectors corresponding to the smallest eigenvalues ​​among all eigenvalues ​​are extracted to form the noise subspace projection matrix.

[0036] Obtain the channel estimation matrix of the multi-load parallel communication system, and perform orthogonal projection processing on the channel estimation matrix onto the noise subspace projection matrix to generate an interference suppression matrix;

[0037] The number of time-domain taps of the space-time joint filter is determined based on the dimension of the interference suppression matrix. Each time-domain tap corresponds to a tap coefficient submatrix. All tap coefficient submatrixes are stacked in time delay order to form the tap coefficient matrix.

[0038] Each symbol vector in the multi-load symbol sequence is concatenated with several adjacent symbol vectors to form an extended symbol vector. The extended symbol vector is then multiplied by the tap coefficient matrix to obtain the filtered output vector.

[0039] Arrange the filtered output vectors corresponding to all symbol vectors in chronological order to obtain the parallel symbol stream after interference suppression.

[0040] As a further aspect of the present invention, residual interference energy calculation processing is performed on the parallel symbol stream after interference suppression to generate residual interference energy values. These residual interference energy values ​​are then fed back to the recursive update step of the improved minimum mean square error algorithm to correct the estimated interference covariance matrix value at the next time step, including:

[0041] Extract the filtered output value of the pilot symbol position from the parallel symbol stream after interference suppression, obtain the real value of the transmitted pilot symbol at the corresponding position, and perform a symbol-by-symbol subtraction operation between the filtered output value and the real value of the transmitted pilot symbol to generate a residual symbol sequence.

[0042] For each residual symbol in the residual symbol sequence, the magnitude squared is calculated, and the sum of the magnitude squared calculation results is divided by the number of residual symbols to obtain the residual interference energy value.

[0043] The residual interference energy value is compared with a preset interference energy threshold value. When the residual interference energy value is greater than the interference energy threshold value, an enhancement update trigger flag is generated.

[0044] Based on the enhanced update trigger flag, the forgetting factor of the recursive least squares algorithm is adjusted to an enhanced forgetting factor value that is less than the current value, and the adjusted forgetting factor value is used for the next time step weight calculation in the exponential weighted moving average processing corresponding to the interference covariance matrix estimate.

[0045] As a further aspect of the present invention, the method further includes adaptive step size adjustment processing on the recursive update step of the improved minimum mean square error algorithm, wherein the adaptive step size adjustment processing includes:

[0046] Extract the error sign between the input symbol vector at the current time and the updated weight vector at the current time, and calculate the moving average magnitude of the error sign;

[0047] The time-varying severity level of the current channel is determined based on the moving average amplitude value, and the time-varying severity level includes slow time-varying level, medium time-varying level and fast time-varying level;

[0048] When the time-varying intensity is classified as slow time-varying, the forgetting factor of the recursive least squares algorithm is set to a large value within the first numerical range. When the time-varying intensity is classified as medium time-varying, the forgetting factor is set to a medium value within the second numerical range. When the time-varying intensity is classified as fast time-varying, the forgetting factor is set to a small value within the third numerical range.

[0049] The set forgetting factor is used as the control parameter for the current iteration step size of the recursive least squares algorithm, and is used to calculate the updated values ​​of the gain vector and the weight vector at the current time.

[0050] As a further aspect of the present invention, the method further includes a step of sparsifying the interference coefficient matrix, wherein the sparsification process includes:

[0051] Iterate through all off-diagonal elements in the interference coefficient matrix at the current time and calculate the magnitude of each off-diagonal element;

[0052] The modulus of each off-diagonal element is compared with a preset sparsification threshold. When the modulus of the off-diagonal element is less than the sparsification threshold, the value of the corresponding off-diagonal element is set to zero. When the modulus of the off-diagonal element is greater than or equal to the sparsification threshold, the original value of the corresponding off-diagonal element is retained.

[0053] The sparsity of the matrix is ​​obtained by calculating the proportion of the total number of non-zero elements in the interference coefficient matrix after zeroing out the total number of elements.

[0054] When the sparsity of the matrix is ​​less than the preset lower limit threshold of sparsity, the sparsification threshold value is gradually reduced and the step of comparing the modulus value of each off-diagonal element with the preset sparsification threshold value is repeated until the sparsity of the matrix is ​​greater than or equal to the lower limit threshold of sparsity, and a sparsification interference coefficient matrix is ​​obtained.

[0055] The sparsed interference coefficient matrix is ​​used to replace the original interference coefficient matrix in subsequent matrix multiplication operations.

[0056] As a further aspect of the present invention, the method further includes performing iterative interference cancellation processing on the interference-suppressed parallel symbol stream, the iterative interference cancellation processing including:

[0057] The parallel symbol stream after interference suppression is used as the current iteration input symbol stream. Hard decision processing is performed on the current iteration input symbol stream to generate a decision symbol sequence.

[0058] The decision symbol sequence is convolved with the estimated interference covariance matrix to generate a reconstructed interference signal sequence.

[0059] Subtract the reconstructed interference signal sequence from the multi-load symbol sequence to generate the interference-cancelled residual symbol stream;

[0060] The residual symbol stream after interference cancellation is used as the new current iteration input symbol stream. Hard decision processing, convolution operation on the decision symbol sequence, and subtraction operation on the multi-load symbol sequence are repeatedly performed on the current iteration input symbol stream until a preset number of iterations is reached or the energy of the reconstructed interference signal sequence is lower than a preset energy threshold. The final interference-cancelled symbol stream is then output as the enhanced interference suppression result.

[0061] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0062] The improved minimum mean square error algorithm transforms the estimation of the interference covariance matrix into a recursive calculation process by integrating the time-varying channel autocorrelation matrix update rule of the recursive least squares algorithm. This avoids the high-dimensional matrix inversion operation in the traditional MMSE algorithm, reduces computational complexity while ensuring estimation accuracy, and enables multi-load parallel communication systems to maintain a stable interference suppression response speed in fast time-varying channel environments.

[0063] The tap coefficient matrix of the space-time joint filter is dynamically constructed based on the real-time updated interference covariance matrix. Combined with the feedback mechanism of residual interference energy value, a closed-loop optimization path of "estimation-filtering-evaluation-correction" is formed, which enables the filter parameters to be adaptively adjusted according to the actual interference suppression effect. This overcomes the performance degradation problem caused by the accumulation of estimation errors in traditional open-loop systems and maintains continuously optimized suppression capability in a dynamically changing interference environment.

[0064] The linkage between residual interference energy calculation and recursive update steps enables dynamic calibration of the interference suppression process. It can autonomously adapt to complex electromagnetic environments without relying on preset channel model parameters, and balances interference suppression accuracy and system resource consumption in multi-load parallel communication scenarios. Attached Figure Description

[0065] Figure 1 This is a flowchart of an adaptive interference suppression method for multi-load parallel communication as described in this invention;

[0066] Figure 2 A flowchart for baseband data frame synchronization extraction;

[0067] Figure 3 The flowchart for updating the weighted moving average estimate. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0069] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0070] See Figure 1 This invention provides an adaptive interference suppression method for multi-payload parallel communication. The method includes: acquiring a baseband sampled data stream from a multi-payload parallel communication system; performing data frame boundary detection and synchronization header extraction on the baseband sampled data stream to eliminate timing and carrier deviations, generating a multi-payload symbol sequence synchronized in time and phase; invoking an improved least mean square error algorithm to estimate the interference matrix of the multi-payload symbol sequence; this improved least mean square error algorithm is based on a recursive least squares algorithm and performs calculations according to the autocorrelation matrix recursive update rule of a time-varying channel, ultimately generating an estimated value of the covariance matrix for the current interference environment; constructing the tap coefficient matrix of a space-time joint filter based on the estimated interference covariance matrix; and performing filtering processing on the aforementioned synchronized multi-payload symbol sequence through the space-time joint filter, thereby jointly suppressing interference in the spatial and temporal domains to obtain an interference-suppressed parallel symbol stream. The residual interference energy is calculated on the parallel symbol stream after interference suppression to generate a residual interference energy value that characterizes the current suppression effect. This value is then fed back into the recursive update step of the improved minimum mean square error algorithm to dynamically correct the estimated value of the interference covariance matrix at the next time step, forming an adaptive closed loop.

[0071] In one embodiment of the present invention, see [reference] Figure 2The baseband sampled data stream from the acquired multi-payload parallel communication system is subjected to matched filtering to generate a matched-filtered baseband complex envelope signal. A sliding cross-correlation operation is then performed between the matched-filtered baseband complex envelope signal and a locally stored frame synchronization sequence to generate a cross-correlation peak sequence. The position of the peak exceeding a preset detection threshold in this cross-correlation peak sequence is detected and determined as the frame start position. Based on the determined frame start position, a complete data frame is extracted from the original baseband sampled data stream, and the synchronization header symbol segment is extracted from a specified position within this complete data frame. The extracted synchronization header symbol segments are compared with a preset synchronization header template for phase error estimation. The process is as follows: The complex value of each synchronization header symbol is sequentially read from the synchronization header symbol segment, and the corresponding standard complex symbol value is read from the synchronization header template. The complex value of each synchronization header symbol is multiplied by the conjugate of its corresponding standard complex symbol value to obtain the complex phase difference at each symbol position. The arctangent of these phase differences is performed to obtain the instantaneous phase error value at each symbol position. The instantaneous phase error values ​​at all symbol positions are arithmetically averaged, and the resulting average phase error value is used as the estimated carrier phase offset of the system. The phase difference between two adjacent symbols in the synchronization header symbol segment and the phase difference between two adjacent standard symbols in the synchronization header template are differentially calculated to obtain the timing error difference value between adjacent symbols. All timing error difference values ​​are summed, and the sum is used as the estimated timing error value of the system. Based on the obtained carrier phase offset estimate and timing error estimate, corresponding phase rotation compensation and timing resampling processing are performed on the data symbol segments in the complete data frame to finally generate the synchronized multi-payload symbol sequence.

[0072] In specific implementation, the process involves acquiring the baseband sampling data stream from a multi-payload parallel communication system, performing data frame boundary detection and synchronization header extraction on the baseband sampling data stream, and generating a synchronized multi-payload symbol sequence. Taking a communication system with four parallel payloads, QPSK modulation, and a frame length of 1024 symbols as an example, the baseband sampling data stream is a complex sequence sampled at twice the symbol rate. Matched filtering is applied to the acquired baseband sampling data stream. The pulse shape of the matched filter is a root-raised cosine roll-off filter with a roll-off factor set to 0.35. After matched filtering, a baseband complex envelope signal with minimal inter-wavelength interference is generated. In some embodiments, the frame synchronization sequence is a pseudo-random binary sequence of length 64. A sliding cross-correlation operation is performed between the matched-filtered baseband complex envelope signal and the locally stored frame synchronization sequence. The sliding window moves one sampling point at a time, generating a cross-correlation peak sequence. The peak position exceeding a preset detection threshold in the cross-correlation peak sequence is used as the frame start position. The detection threshold is set to 0.8 times the maximum peak amplitude. Based on the frame start position, a complete data frame is extracted from the original baseband sampled data stream, with a length of 2048 sampling points. Starting from the 16th sampling point after the start position of the complete data frame, 128 consecutive sampling points are extracted to form the synchronization header symbol segment.

[0073] In practice, the synchronization header template stores the theoretical complex values ​​of the standard synchronization header symbols at the corresponding sampling points after shaping and filtering. Phase error estimation is performed between the synchronization header symbol segment and the preset synchronization header template to generate carrier phase offset and timing error estimates. The complex value of each synchronization header symbol is sequentially read from the synchronization header symbol segment, and the corresponding standard complex symbol value is read from the synchronization header template. A complex multiplication operation is performed between the complex value of each synchronization header symbol and the conjugate of the corresponding standard complex symbol value to obtain the complex phase difference at each symbol position.

[0074] In some embodiments, the arctangent of the phase difference is performed to obtain the instantaneous phase error value at each symbol position. The arithmetic mean of the instantaneous phase error values ​​at all symbol positions is then calculated, and the calculation process can be expressed by the following formula:

[0075] ;

[0076] in: This represents the estimated carrier phase offset. This indicates the total number of symbols in the synchronization header symbol field. This represents the complex value of the k-th symbol in the synchronization header symbol field. This represents the conjugate of the k-th standard complex number sign value in the synchronization header template. The function represents taking the real part of a complex number. The function represents taking the imaginary part of a complex number. The function represents the arctangent operation. The calculated average phase error value is used as the carrier phase offset estimate. Optionally, when calculating the timing error estimate, the phase difference between two adjacent symbols in the synchronization header symbol segment and the phase difference between two adjacent standard symbols in the synchronization header template are differentially analyzed. The phase difference between the k-th symbol and the (k+1)-th symbol in the synchronization header symbol segment is calculated. Calculate the phase difference between the k-th standard symbol and the (k+1)-th standard symbol at the corresponding position in the synchronization header template:

[0077] ;

[0078] in: The function represents the calculation of the phase angle of a complex number. The timing error differential value is obtained by performing a difference operation on these two phase differences. All timing error differential values ​​are summed up. The summation result is... As an estimate of timing error.

[0079] In practice, compensation processing is performed on the data symbol segments in the complete data frame based on the obtained carrier phase offset estimate and timing error estimate. The carrier phase offset estimate is used to perform phase rotation on each sampling point in the data symbol segment; the carrier phase offset estimate with a negative rotation angle is used. The timing error estimate is used to control the resampling filter, performing timing resampling on the data symbol segment and adjusting the timing phase of the sampling points to eliminate timing deviations between symbols. After phase rotation and timing resampling, a multi-payload symbol sequence is generated, synchronized with both time and carrier phase. Each time point in the multi-payload symbol sequence contains four parallel complex symbol values, corresponding to four parallel payloads. It can be understood that the root-raised cosine filter coefficients used in the matched filtering process are pre-designed and stored. In the sliding cross-correlation operation, the locally stored frame synchronization sequence is prior information known to the receiver. The phase error estimation process relies on the comparison between the synchronization header symbol segment and the synchronization header template, which is a standard sequence pre-agreed upon by the transmitter and receiver. Optionally, in frame start position detection, if no peak exceeding the threshold is detected in the cross-correlation peak sequence, the sliding window is expanded or the detection threshold is adjusted before re-detection until the frame start position is successfully detected. In phase error estimation, if there are severely interfered symbols in the synchronization header symbol segment, a robust averaging method after removing outliers can be used when calculating the sum of the average phase error and timing error differential values. It can be understood that generating a synchronized multi-load symbol sequence is a prerequisite for subsequent interference estimation and suppression. The accuracy of the synchronization operation directly affects the effectiveness of the interference covariance matrix estimate, and the accuracy of the carrier phase offset estimate and timing error estimate determines the alignment quality of the multi-load symbol sequence, thus affecting the filtering performance of the space-time joint filter.

[0080] In one embodiment of the present invention, initializing the parameters of the recursive least squares algorithm includes setting the forgetting factor value, setting the initial value of the inverse correlation matrix, and setting the initial value of the weight vector. See also... Figure 3 The synchronized multi-load symbol sequence is sequentially input into the iterative update steps of the recursive least squares algorithm in chronological order. Each iterative update step includes: calculating the gain vector based on the input symbol vector at the current time and the inverse correlation matrix stored at the previous time; calculating the updated weight vector value at the current time based on the weight vector at the previous time, the calculated gain vector, and the expected symbol value at the current time; and updating the inverse correlation matrix value at the current time based on the inverse correlation matrix at the previous time and the calculated gain vector. After each iterative update step, off-diagonal elements are extracted from the updated weight vector value at the current time, and all off-diagonal elements are arranged according to load index to form the interference coefficient matrix at the current time. The interference coefficient matrix at the current time is then multiplied by its conjugate transpose to generate the instantaneous interference covariance matrix at the current time. Based on the forgetting factor, an exponentially weighted moving average is applied to the interference covariance matrix at historical moments and the instantaneous interference covariance matrix at the current moment. Specifically: after each iteration update step of the recursive least squares algorithm, the instantaneous interference covariance matrix calculated at the current moment is obtained; the estimated value of the interference covariance matrix generated and stored in the previous iteration update step is obtained as the interference covariance matrix at historical moments; the currently valid forgetting factor value is read from the parameters of the recursive least squares algorithm, and the first weighting coefficient and the second weighting coefficient are calculated, where the first weighting coefficient is equal to the forgetting factor value, and the second weighting coefficient is equal to one minus the forgetting factor value; the interference covariance matrix at historical moments is multiplied by the first weighting coefficient, and the instantaneous interference covariance matrix at the current moment is multiplied by the second weighting coefficient; the results of the two multiplication operations are matrix added to obtain the estimated value of the interference covariance matrix at the current moment; finally, the estimated value of the interference covariance matrix at the current moment is stored and used as the interference covariance matrix at historical moments in the next iteration update step.

[0081] In practical implementation, an improved least mean square error (LMS) algorithm is used to estimate the interference matrix of the multi-payload symbol sequence. This improved LMS algorithm is based on a recursive least squares algorithm and uses the autocorrelation matrix update rule of the time-varying channel to generate the estimated interference covariance matrix. In a communication scenario with four parallel transmissions, the synchronized multi-payload symbol sequence at each time step k can be represented as a four-dimensional complex input symbol vector:

[0082] ;

[0083] in: This represents the sign of the i-th load at time k. Initialize the parameters of the recursive least squares algorithm and set the forgetting factor value. Set the initial value of the inverse correlation matrix to 0.99. Multiply a 4x4 identity matrix by a regularization constant δ (e.g., δ=0.01) and set the initial value of the weight vector. It is a 4x1 zero vector.

[0084] In some embodiments, the multi-load symbol sequence is input into the iterative update step of the recursive least squares algorithm in time order k=1,2,3,… For each time k, the iterative update step includes adjusting the input symbol vector based on the current time. inverse correlation matrix with the previous time step Calculate the gain vector The calculation method is as follows:

[0085] ;

[0086] Among them: superscript This represents the conjugate transpose. Based on the weight vector from the previous time step. The calculated gain vector and the expected sign value at the current moment. Calculate the updated weight vector value at the current time step. The calculation method is as follows:

[0087] ;

[0088] in: It is a known pilot or decision feedback symbol. Indicates conjugate. Based on the inverse correlation matrix of the previous time step. and gain vector Calculate the updated value of the inverse correlation matrix at the current time. The calculation method is as follows .

[0089] In practice, after each iterative update step, the updated weight vector value at the current moment is extracted. The off-diagonal elements in the weight vector. It is a 4x1 vector that does not directly contain off-diagonal elements. Here, it actually refers to constructing an interference relation matrix using the relationship between the weight vector and the input vector. One implementation method is to use the equivalent channel or interference response matrix estimated at the current time. Consider it as a function related to the weight vector, for example, through Obtained in the form of, among which It is the identity matrix. It is the guide vector, and then from Extract the off-diagonal elements. Arrange all off-diagonal elements according to their load indices to form a 4x4 interference coefficient matrix for the current time step. ,matrix off-diagonal elements (i≠j) represents the estimated interference coupling coefficient between the j-th load and the i-th load. The interference coefficient matrix at the current time... Its conjugate transpose Perform matrix multiplication to generate the instantaneous disturbance covariance matrix at the current moment. .

[0090] Optionally, an exponentially weighted moving average is applied to the instantaneous disturbance covariance matrices of historical moments and the current moment based on a forgetting factor to generate a smoothed estimate of the disturbance covariance matrix. After each iteration update step of the recursive least squares algorithm, the instantaneous disturbance covariance matrix calculated at the current time is obtained. Obtain the estimated value of the disturbance covariance matrix generated and stored in the previous iteration update step. This serves as the disturbance covariance matrix for historical moments. The currently valid forgetting factor values ​​are read from the parameters of the recursive least squares algorithm. The first weighting coefficient and the second weighting coefficient are calculated, where the first weighting coefficient is equal to the forgetting factor value. The second weighting coefficient is equal to one minus the forgetting factor value, i.e. The process of exponentially weighted moving average can be represented by the following formula:

[0091] ;

[0092] in: This represents the estimated value of the disturbance covariance matrix at time k. This represents the estimated value of the disturbance covariance matrix at time k-1. This represents the instantaneous disturbance covariance matrix at time k. It is the forgetting factor, with a value ranging from 0.9 to 1. It is the second weighting coefficient. The disturbance covariance matrix at historical moments... With the first weighting coefficient Perform scalar multiplication to obtain the instantaneous disturbance covariance matrix at the current moment. With the second weighting coefficient Perform scalar multiplication, then add the results of the two multiplications together to obtain the estimated value of the disturbance covariance matrix at the current time. The estimated value of the disturbance covariance matrix at the current time. This is stored and used as the disturbance covariance matrix for computation in the next iteration update step, representing a historical moment. It's understandable that the initialization parameters of the recursive least squares algorithm affect the convergence speed; the initial value of the inverse correlation matrix... Setting it to a multiple of the identity matrix is ​​a common practice. Expected sign value Known pilot symbols are used during the training phase, and decision guidance can be used in the data segment to obtain the interference coefficient matrix. The construction method reflects the mutual interference relationship between parallel loads.

[0093] In some embodiments, forgetting factor The value can be adjusted according to the channel's rate of change; a slower-changing channel can use a larger value. To obtain a smoother estimate, a smaller value can be used for channels that change more rapidly. The values ​​are rapidly tracked for change. The exponentially weighted moving average effectively smooths out fluctuations in instantaneous estimates, providing a more robust estimate of the disturbance covariance matrix. This is used for calculating the coefficients of the subsequent space-time joint filter. Optionally, the instantaneous interference covariance matrix... The calculation may require adding a small regularization term, such as To prevent matrix singularities, where It is a very small positive number. Estimated value of the disturbance covariance matrix. The storage can use a double-buffered or circular queue structure to ensure efficient data access during continuous processing.

[0094] In one embodiment of the present invention, the estimated interference covariance matrix is ​​subjected to eigenvalue decomposition, and the eigenvectors corresponding to the smallest eigenvalues ​​are extracted. These eigenvectors form the noise subspace projection matrix. After each iterative update step, an interference coefficient matrix is ​​constructed based on the updated weight vector values ​​at the current time. The off-diagonal elements in the interference coefficient matrix are extracted, and all off-diagonal elements are arranged according to the load index to form the interference coefficient matrix at the current time. The channel estimation matrix of the multi-load parallel communication system is obtained, and the channel estimation matrix is ​​orthogonally projected onto the noise subspace projection matrix to generate an interference suppression matrix for suppressing interference. The number of time-domain taps required for the space-time joint filter is determined according to the dimension of the interference suppression matrix. Each time-domain tap corresponds to a tap coefficient submatrix, and all tap coefficient submatrices are stacked in time delay order to form a complete tap coefficient matrix. Each symbol vector in the multi-load symbol sequence is concatenated with several adjacent symbol vectors to form an extended symbol vector. The extended symbol vector is multiplied by the constructed tap coefficient matrix to obtain the filtered output vector of the current symbol. Arrange the filtered output vectors corresponding to all symbol vectors in their original time order to obtain the interference-suppressed parallel symbol stream. Perform residual interference energy calculation on the interference-suppressed parallel symbol stream. The process is as follows: extract the filtered output value at the pilot symbol position from the interference-suppressed parallel symbol stream and obtain the actual value of the transmitted pilot symbol at the corresponding position. Perform symbol-by-symbol subtraction on the filtered output value at each position and the actual value of the transmitted pilot symbol to generate a residual symbol sequence. Calculate the square of the modulus length for each residual symbol in the residual symbol sequence. Sum the results of all modulus length square calculations and divide by the total number of residual symbols to obtain the residual interference energy value. Compare this residual interference energy value with a preset interference energy threshold. When the residual interference energy value is greater than the threshold, generate an enhancement update trigger flag. Based on this enhancement update trigger flag, adjust the forgetting factor value used in the recursive least squares algorithm to a value less than the current value. Use the adjusted forgetting factor value for the next time step weight calculation in the exponential weighted moving average processing corresponding to the interference covariance matrix estimate.

[0095] In practical implementation, the tap coefficient matrix of the space-time joint filter is constructed based on the estimated interference covariance matrix. The multi-load symbol sequence is then filtered by the space-time joint filter to obtain the interference-suppressed parallel symbol stream. Residual interference energy calculation and feedback correction are then performed on the interference-suppressed parallel symbol stream. Assuming the multi-load parallel communication system contains four parallel data streams, the estimated interference covariance matrix... It is a 4x4 complex numerical matrix. The estimated value of the disturbance covariance matrix... Eigenvalue decomposition is performed, and the eigenvalue decomposition is expressed as:

[0096] ;

[0097] in: It is a unitary matrix composed of eigenvectors. It is a diagonal matrix composed of eigenvalues. The eigenvectors corresponding to the smallest eigenvalues ​​are extracted from all eigenvalues ​​to form the projection matrix of the noise subspace. In practice, after the eigenvalues ​​are sorted in ascending order, the eigenvectors corresponding to the two smallest eigenvalues ​​are selected as the noise subspace. Therefore, the projection matrix of the noise subspace is... It is a 4x2 matrix.

[0098] In some embodiments, the channel estimation matrix of a multi-payload parallel communication system is obtained. Channel estimation matrix It is a 4x4 matrix reflecting the transmission channels between the payloads. The channel estimation matrix... Projection matrix to the noise subspace Perform orthogonal projection processing to generate an interference suppression matrix. The projection operation can be expressed as:

[0099] ;

[0100] in: It is a 4x4 identity matrix. Based on the interference suppression matrix... The dimension determines the number of time-domain taps in the space-time joint filter. Number of time-domain taps It is necessary to adapt to the maximum delay spread characteristics of the channel, and the specific determination rules are as follows: Based on the channel delay spread: number of time-domain taps Should meet ,in Maximum channel delay spread (in seconds). Signal bandwidth (unit: Hz) Indicates rounding up; based on the number of antenna arrays: when the delay extension is unknown or a simplified design is required, it can be based on the number of antenna arrays. set up ( (This indicates rounding up).

[0101] In this embodiment, the multi-payload parallel communication system includes four parallel data streams (i.e., the number of antenna arrays). ), signal bandwidth Maximum channel delay spread Calculated according to rule 1 ;

[0102] Calculated according to the rules

[0103] ;

[0104] To balance interference suppression performance with computational complexity, the actual number of time-domain taps is set. (One tap is added compared to the theoretical value to cover edge delay components), each time-domain tap corresponds to one tap coefficient submatrix All tap coefficient submatrices are stacked in time delay order to form the total tap coefficient matrix. , It is A matrix (3 taps * 4 rows x 4 columns).

[0105] Based on the structure of the space-time filter, an extended symbol vector of length 12 is constructed. This vector is composed of the emission symbols from the current time step and the two adjacent time steps before and after it, and its specific components are as follows:

[0106] ;

[0107] in, For the previous moment Emit symbol vector, For the current moment Emit symbol vector, This is the preset / estimated sign vector for the next time step. The three are concatenated to form... An extended symbol vector of dimension is used as the input to the space-time filter.

[0108] Optionally, for the parallel symbol stream after interference suppression Perform residual interference energy calculation processing to generate residual interference energy values. From the interference-suppressed parallel symbol stream... The filtered output value is extracted from the pilot symbol positions. Assuming each data block contains N_p pilot symbols, their position index set is... Obtain the actual value of the transmitted pilot symbol at the corresponding location. Each of them It is a known 4-dimensional pilot vector. The filtered output value... Compared with the actual value of the transmitted pilot symbol Perform sign-by-sign subtraction to generate a residual sign sequence. i=1,2,..., For each residual symbol in the residual symbol sequence Perform modulus square calculation, sum the results of the modulus square calculation, and divide by the number of residual signs to obtain the residual interference energy value. Residual interference energy value The calculation formula is: ;

[0109] in: This indicates the residual interference energy value. Indicates the total number of pilot symbols. Represents the residual symbol vector (4-dimensional) at the i-th pilot position. This represents calculating the square of the L2 norm of a vector (i.e., the sum of the squares of the moduli of each element of the vector). This can be understood as the residual interference energy value. This reflects the power level of the remaining interference and noise after passing through the space-time joint filter, see Table 1.

[0110] Table 1: Noise Subspace Feature Vector Selection Table

[0111]

[0112] In some embodiments, the calculated residual interference energy value With a preset interference energy threshold Compare the residual interference energy values. Greater than the interference energy threshold At this time, an enhancement update trigger flag is generated. Based on this enhancement update trigger flag, the forgetting factor value of the recursive least squares algorithm is changed from the current value. Adjust to a value lower than the current value of the enhanced forgetting factor. For example, The value was adjusted from 0.99 to 0.95. (This is the adjusted forgetting factor value.) The weights used in the next time step of the exponentially weighted moving average processing of the interference covariance matrix estimate are calculated, i.e., in the next iteration. Use a new forgetting factor.

[0113] It is understandable that the construction of the noise subspace projection matrix relies on the eigenvalue decomposition of the interference covariance matrix estimate, and selecting the eigenvector corresponding to the smallest eigenvalue can effectively characterize the interference noise subspace. The space-time joint filter suppresses interference by projecting the channel response onto the orthogonal complement space of the noise subspace. The calculation of residual interference energy utilizes known pilot symbols, providing quantitative feedback information for evaluating the interference suppression effect and adjusting the estimation algorithm parameters.

[0114] Optional, interference energy threshold The settings can be configured based on the system noise floor and acceptable performance margin. The enhancement update trigger flag can be set to a Boolean variable, set to "true" when a condition is met. The forgetting factor can be adjusted incrementally; for example, multiplying the forgetting factor by a factor less than 1 each time residual energy is detected to exceed the limit. Tap coefficient matrix. It needs to be recalculated periodically based on channel changes or updates to the interference covariance matrix.

[0115] In one embodiment of the present invention, an adaptive step size adjustment process is performed in the recursive update step. The error symbol between the input symbol vector at the current time and the updated weight vector value at the current time is extracted, and the moving average magnitude of the error symbol is calculated. Based on the calculated moving average magnitude, the time-varying severity level of the current channel is determined. This time-varying severity level includes slow time-varying, medium time-varying, and fast time-varying levels. When the time-varying severity level is determined to be slow, the forgetting factor of the recursive least squares algorithm is set to a large value within a first numerical interval; when it is determined to be medium, the forgetting factor is set to a middle value within a second numerical interval; and when it is determined to be fast, the forgetting factor is set to a small value within a third numerical interval. The set forgetting factor is used as a control parameter for the current iteration step size of the recursive least squares algorithm to calculate the gain vector and the updated weight vector value at the current time. After obtaining the interference coefficient matrix, it is sparsified. All off-diagonal elements in the interference coefficient matrix at the current time are traversed, and the magnitude of each off-diagonal element is calculated. The magnitude of each off-diagonal element is compared with a preset sparsity threshold. If the magnitude of an off-diagonal element is less than the threshold, its value is set to zero; if the magnitude is greater than or equal to the threshold, its original value is retained. The proportion of non-zero elements to the total number of elements in the interference coefficient matrix after this zeroing process is calculated to obtain the matrix sparsity. When the calculated matrix sparsity is less than a preset lower sparsity threshold, the sparsity threshold is gradually lowered, and the comparison of the magnitude of each off-diagonal element with the threshold is repeated until the matrix sparsity is greater than or equal to the lower threshold, thus obtaining the sparsified interference coefficient matrix. This sparsified interference coefficient matrix replaces the original interference coefficient matrix for subsequent matrix multiplication operations to generate the instantaneous interference covariance matrix.

[0116] In practical implementation, the recursive update step of the improved least mean square error algorithm undergoes adaptive step size adjustment, and the interference coefficient matrix is ​​sparsified. The adaptive step size adjustment dynamically adjusts the forgetting factor based on the current time-varying severity of the channel, while the sparsification process reduces the complexity of the interference coefficient matrix. Consider a communication system with four parallel payloads; the recursive least squares algorithm generates error symbols at each time step k. Error symbol It is the expected sign value at the current moment. The difference between the filter output and the output, i.e. Extract the input symbol vector at the current time step. Updated weight vector values ​​at the current time Error sign between Calculate the error sign The moving average amplitude value Moving average amplitude value The calculation formula is: ;

[0117] in: This represents the moving average amplitude at time k. It is a smoothing factor between 0 and 1 (e.g., 0.95). Indicates the error symbol The magnitude of the moving average. It reflects the average level of recent errors, and its magnitude is related to the drasticness of channel changes.

[0118] In some embodiments, based on the moving average amplitude value The time-varying severity level of the current channel is determined, including slow, medium, and fast time-varying levels. Two amplitude thresholds are set. and (For example , When the moving average amplitude value Less than or equal to the threshold When the time-varying intensity is determined, it is classified as a slow time-varying level; when the moving average amplitude value Greater than the threshold And less than or equal to the threshold When the time-varying intensity is determined to be moderate, the moving average amplitude value is... Greater than the threshold At that time, the severity of the time-varying phenomenon was classified as rapid time-varying. Different classifications correspond to different forgetting factors. The numerical range is set such that when the degree of time-varying drasticness is classified as slow time-varying, the forgetting factor of the recursive least squares algorithm is applied. Set to a large value within the first numerical range, such as a value within the range [0.995, 0.999]; when the time-varying intensity is classified as moderate, the forgetting factor will be... Set to the middle value within the second numerical range, for example, a value within the range [0.98, 0.995]; when the time-varying severity is classified as rapid time-varying, the forgetting factor is... Set it to a small value within the third numerical range, for example, a value within the range [0.95, 0.98]. The resulting forgetting factor will be... As a control parameter for the current iteration step size of the recursive least squares algorithm, it is used to calculate the gain vector. and the updated weight vector value at the current time. See Table 2.

[0119] Table 2: Correspondence between Time-Variation Severity Grading and Forgetting Factor Settings

[0120]

[0121] In practical implementation, after obtaining the interference coefficient matrix, the step is to perform sparsification processing on the interference coefficient matrix. The interference coefficient matrix at the current time step is obtained from the iteration of the recursive least squares algorithm. It is a 4x4 complex matrix, where the diagonal elements typically represent the desired signal components rather than interference. Sparsity processing mainly targets the off-diagonal elements. Iterate through the interference coefficient matrix at the current time step. All non-diagonal elements (in ), calculate each off-diagonal element modulus The modulus of each off-diagonal element. With a preset sparsity threshold Comparison, sparsification threshold The initial value can be set to 0.1. When the modulus of the off-diagonal elements... Less than the sparsity threshold When, the corresponding off-diagonal elements will be... The value is set to zero; when the modulus of the off-diagonal element is zero. Greater than or equal to the sparsity threshold value When this happens, retain the corresponding off-diagonal element. The original value.

[0122] Optionally, the matrix sparsity can be obtained by calculating the proportion of the total number of non-zero elements in the interference coefficient matrix after zeroing out to the total number of elements. For a 4x4 matrix, the total number of elements is 16, and the number of non-zero elements is... Then the sparsity of the matrix When the matrix sparsity Less than a preset sparsity lower threshold Time (e.g.) This indicates that an overly sparse matrix may lose effective interference information, and the sparsity threshold needs to be gradually reduced. And it repeatedly performs the step of comparing the modulus of each off-diagonal element with a preset sparsity threshold. For example, comparing the sparsity threshold... Reduce the threshold from 0.1 to 0.05, and then re-traverse the matrix using the new threshold value. The off-diagonal elements are compared and zeroed out, and the matrix sparsity is calculated again. until the matrix sparsity Greater than or equal to the sparsity lower limit threshold The final sparsed interference coefficient matrix is ​​obtained. The sparse interference coefficient matrix is ​​then used. Replace the original interference coefficient matrix Used for subsequent matrix multiplication operations to generate the instantaneous disturbance covariance matrix. .

[0123] It is understandable that adaptive step size adjustment senses the channel change rate by monitoring the error magnitude and adjusts the forgetting factor accordingly. A larger forgetting factor indicates a longer memory period and smoother tracking of slowly changing channels, while a smaller forgetting factor indicates a more sensitive algorithm and stronger ability to track rapidly changing channels. Sparse processing, by filtering out interference coefficients with small magnitudes, can reduce matrix operation complexity and errors caused by estimation noise. In some embodiments, the moving average magnitude value... The initialization can be set to an empirical value or estimated from the initial training sequence. The threshold for grading the time-varying severity. and The calibration setting can be made according to the actual Doppler range of the channel during system operation. Sparsity threshold value. The initial values ​​and adjustment step sizes can be set according to the typical intensity of the disturbance. Optionally, sparsification can be performed after each iteration update, or it can be performed once every few iterations to balance performance and complexity. Matrix sparsity The calculation can consider only the off-diagonal elements, i.e., the total number of elements is 12 (a 4x4 matrix minus 4 diagonal elements), and the sparsity lower bound threshold is... The corresponding adjustments are also made. Adaptive step size adjustment and sparsification can be enabled independently or used in combination, both working together to improve the interference estimation algorithm.

[0124] In one embodiment of the present invention, the parallel symbol stream after interference suppression output from the space-time joint filter is used as the current iteration input symbol stream. Hard decision processing is performed on this current iteration input symbol stream to generate a decision symbol sequence. This decision symbol sequence is then convolved with the latest estimated interference covariance matrix to generate a reconstructed interference signal sequence. The discrete circular convolution is performed according to the following formula: ;

[0125] In the formula, For discrete-time indexing, This represents the number of time-domain taps in the space-time filter. To estimate the interference covariance matrix in time delay Discrete sampled values ​​at the location, The complex conjugate and shift of the decision symbol sequence The value of the bit is calculated. The reconstructed interference signal sequence is subtracted from the original multi-load symbol sequence to generate a residual symbol stream after interference cancellation. This residual symbol stream is used as the new input symbol stream for the current iteration. Hard decision processing on the current input symbol stream, convolution operation on the decision symbol sequence, and subtraction operation on the original multi-load symbol sequence are repeatedly performed, forming an iterative loop. This loop continues until a preset maximum number of iterations is reached, or the energy of the reconstructed interference signal sequence generated in the current iteration is lower than a preset energy threshold. Finally, the interference-cancelled symbol stream obtained in the last iteration is output as the enhanced interference suppression result.

[0126] In practice, the process of performing iterative interference cancellation processing on the interference-suppressed parallel symbol stream involves using the interference-suppressed parallel symbol stream output by the space-time joint filter as the initial input symbol stream for the current iteration. ,in Indicates the symbol time index, superscript This represents the 0th iteration (i.e., the initial input). Taking a system with four parallel data streams as an example, the current iteration input symbol stream... In each iteration Each time is a four-dimensional complex vector sequence. For the current iteration, the input symbol stream... Hard decision processing is performed, which maps each complex symbol value to the nearest constellation point according to the modulation scheme (e.g., QPSK) used in the symbol stream, generating a decision symbol sequence. , decision symbol sequence The dimension is the same as the current iteration's input symbol stream.

[0127] In some embodiments, the decision symbol sequence Compared with the latest disturbance covariance matrix estimate obtained from the improved minimum mean square error algorithm Perform convolution operations to generate a reconstructed interference signal sequence, and estimate the interference covariance matrix. It is a 4x4 matrix that reflects the interference characteristics between loads. Reconstructing the interference signal sequence. The calculation involves estimating the interference covariance matrix. With the sequence of judgment symbols Linear convolution can be specifically represented as a weighted sum at multiple symbol times. A specific convolution operation formula can be expressed as:

[0128] ;

[0129] in: Indicates the first Next iteration, time step The reconstructed interference signal vector (4-dimensional). Indicates time delay The estimated value of the disturbance covariance matrix (4x4 matrix) on the surface. It takes into account the length of interference memory. It is the first Next iteration, time step The decision symbol vector (4-dimensional). The summation symbol represents the time delay over all considerations. from arrive Accumulate the results. In practice, Possibly through the This can be achieved through transformation or directly derived from channel estimation. From the original multi-payload symbol sequence... Subtract the reconstructed interference signal sequence Generate the residual symbol stream after interference cancellation. .

[0130] In practical implementation, the residual symbol stream after interference removal will be... As the new current iteration input symbol stream The process involves repeatedly performing hard decision processing on the current input symbol stream, convolution operations on the decision symbol sequence, and subtraction operations on the multi-load symbol sequence, forming an iterative loop. The preset upper limit for the number of iterations is denoted as... For example, it can be set to 3. The reconstructed interference signal sequence is calculated after each iteration. energy ,energy It can be calculated The energy threshold is obtained by summing the squares of the moduli of all symbols within a data block. The preset energy threshold is denoted as... The iterative loop continues until the current iteration number is reached. Reaching the preset maximum number of iterations Or the energy of the reconstructed interference signal sequence generated in the current iteration. Below the preset energy threshold The iteration loop terminates when any condition is met, and outputs the symbol stream after interference removal obtained from the last iteration. As a result of enhanced interference suppression.

[0131] It is understandable that iterative interference cancellation utilizes the symbols from the initial decision to reconstruct and subtract residual interference, potentially further improving interference suppression performance. The initial input symbol stream for the current iteration... This is the output of the space-time joint filter, which has already undergone one round of interference suppression. The accuracy of hard decision processing affects the quality of the reconstructed interference signal, and it is more reliable at high signal-to-noise ratios. Optional, a preset upper limit on the number of iterations. It can be set based on the marginal effect of system processing capacity and performance improvement. Preset energy threshold value. This can be set based on the system noise power or acceptable residual interference level. The interference covariance matrix estimate used in the convolution operation. The decision symbol sequence can remain unchanged during iteration, or it can be updated after each iteration using a new residual symbol stream. Hard decision or soft decision (such as using soft information) can be used to improve reconstruction accuracy. In some embodiments, the interference signal sequence is reconstructed. The computation can be performed in parallel for all symbols of a data block to improve processing speed. The residual symbol stream after interference cancellation. Before being used as input for the next iteration, additional linear filtering can be applied to further whiten the noise. The termination condition for the iteration can be an "OR" relationship between two conditions, meaning that the iteration stops when either one is satisfied, or an "AND" relationship, meaning that the iteration count must be reached and the energy must be below a threshold simultaneously before stopping. The "OR" relationship is usually used to avoid unnecessary iterations.

[0132] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An adaptive interference suppression method for multi-load parallel communication, characterized in that, The method includes: Acquire the baseband sampled data stream in a multi-payload parallel communication system, perform data frame boundary detection and synchronization header extraction on the baseband sampled data stream, and generate a synchronized multi-payload symbol sequence; An improved least mean square error algorithm is invoked to perform interference matrix estimation on the multi-load symbol sequence. The improved least mean square error algorithm is based on the recursive least squares algorithm and generates an estimated value of the interference covariance matrix according to the autocorrelation matrix update rule of the time-varying channel. Based on the estimated value of the interference covariance matrix, a tap coefficient matrix of a space-time joint filter is constructed. The multi-load symbol sequence is then filtered by the space-time joint filter to obtain a parallel symbol stream with interference suppression. The residual interference energy is calculated on the parallel symbol stream after interference suppression to generate a residual interference energy value. The residual interference energy value is then fed back to the recursive update step of the improved minimum mean square error algorithm to correct the estimated value of the interference covariance matrix at the next time step.

2. The adaptive interference suppression method for multi-load parallel communication according to claim 1, characterized in that, The process of acquiring the baseband sampled data stream in a multi-payload parallel communication system, performing data frame boundary detection and synchronization header extraction on the baseband sampled data stream, and generating a synchronized multi-payload symbol sequence includes: The baseband sampled data stream is subjected to matched filtering to generate a matched-filtered baseband complex envelope signal; The matched-filtered baseband complex envelope signal is subjected to sliding cross-correlation operation with the locally stored frame synchronization sequence to generate a cross-correlation peak sequence. The peak position in the cross-correlation peak sequence that exceeds the detection threshold is used as the frame start position. Based on the frame start position, a complete data frame is extracted from the baseband sampled data stream, and a synchronization header symbol segment is extracted from a specified position of the complete data frame; The synchronization header symbol segment is compared with a preset synchronization header template to perform phase error estimation processing, generating carrier phase offset estimation value and timing error estimation value; Based on the estimated carrier phase offset and the estimated timing error, the data symbol segments in the complete data frame are subjected to phase rotation and timing resampling to generate the synchronized multi-payload symbol sequence.

3. The adaptive interference suppression method for multi-load parallel communication according to claim 2, characterized in that, The synchronization header symbol segment is compared with a preset synchronization header template to perform phase error estimation processing, generating carrier phase offset estimates and timing error estimates, including: The complex value of each synchronization header symbol is read sequentially from the synchronization header symbol segment, and the standard complex symbol value at the corresponding position is read from the synchronization header template; The complex value of each synchronization header symbol is multiplied by the conjugate of the corresponding standard complex symbol value to obtain the complex phase difference of each symbol position; Perform an arctangent operation on the complex phase difference to obtain the instantaneous phase error value at each symbol position; The instantaneous phase error values ​​of all symbol positions are arithmetically averaged to obtain the average phase error value, which is then used as the estimated carrier phase offset value. The timing error difference between adjacent symbols is obtained by performing a differential operation between the phase difference between two adjacent symbols in the synchronization header symbol segment and the phase difference between two adjacent standard symbols in the synchronization header template. The sum of all timing error difference values ​​is used as the estimated timing error value.

4. The adaptive interference suppression method for multi-load parallel communication according to claim 1, characterized in that, The improved minimum mean square error algorithm is invoked to perform interference matrix estimation on the multi-load symbol sequence. This improved minimum mean square error algorithm, based on a recursive least squares algorithm and following the autocorrelation matrix update rule of the time-varying channel, generates an estimated interference covariance matrix, including: Initialize the forgetting factor, the inverse correlation matrix, and the weight vector of the recursive least squares algorithm; The multi-load symbol sequence is sequentially input into the iterative update step of the recursive least squares algorithm in chronological order. The iterative update step includes: calculating the gain vector based on the input symbol vector at the current time and the inverse correlation matrix at the previous time; calculating the updated value of the weight vector at the current time based on the weight vector at the previous time, the gain vector, and the expected symbol value at the current time; and calculating the updated value of the inverse correlation matrix at the current time based on the inverse correlation matrix at the previous time and the gain vector. After each iterative update step is completed, an interference coefficient matrix is ​​constructed based on the weight vector update value at the current time. The off-diagonal elements in the interference coefficient matrix are extracted, and all off-diagonal elements are arranged according to the load index to form the interference coefficient matrix at the current time. Perform matrix multiplication on the current disturbance coefficient matrix and its conjugate transpose matrix to generate the current instantaneous disturbance covariance matrix. Based on the forgetting factor, an exponentially weighted moving average is applied to the instantaneous disturbance covariance matrix at historical times and the instantaneous disturbance covariance matrix at the current time to generate an estimated value of the disturbance covariance matrix.

5. The adaptive interference suppression method for multi-load parallel communication according to claim 4, characterized in that, Based on the forgetting factor, an exponentially weighted moving average is applied to the instantaneous disturbance covariance matrix at historical times and the instantaneous disturbance covariance matrix at the current time to generate an estimate of the disturbance covariance matrix, including: After each iteration update step of the recursive least squares algorithm is completed, the instantaneous disturbance covariance matrix at the current moment is obtained; Obtain the estimated value of the disturbance covariance matrix generated and stored in the previous iteration update step, and use it as the disturbance covariance matrix at the historical moment. Read the currently valid forgetting factor value from the parameters of the recursive least squares algorithm, and calculate the first weighting coefficient and the second weighting coefficient, wherein the first weighting coefficient is equal to the forgetting factor value, and the second weighting coefficient is equal to one minus the forgetting factor value; The interference covariance matrix at the historical moment is multiplied by the first weighting coefficient, and the instantaneous interference covariance matrix at the current moment is multiplied by the second weighting coefficient. The results of the two multiplication operations are then added to obtain the estimated value of the interference covariance matrix at the current moment. The estimated value of the disturbance covariance matrix at the current moment is stored and used as the disturbance covariance matrix at the historical moment in the next iteration update step.

6. The adaptive interference suppression method for multi-load parallel communication according to claim 1, characterized in that, Based on the estimated interference covariance matrix, a tap coefficient matrix for a space-time joint filter is constructed. The multi-load symbol sequence is then filtered through the space-time joint filter to obtain a parallel symbol stream with interference suppression, including: The estimated value of the interference covariance matrix is ​​subjected to eigenvalue decomposition, and the eigenvectors corresponding to the smallest eigenvalues ​​among all eigenvalues ​​are extracted to form the noise subspace projection matrix. Obtain the channel estimation matrix of the multi-load parallel communication system, and perform orthogonal projection processing on the channel estimation matrix onto the noise subspace projection matrix to generate an interference suppression matrix; The number of time-domain taps of the space-time joint filter is determined based on the dimension of the interference suppression matrix. Each time-domain tap corresponds to a tap coefficient submatrix. All tap coefficient submatrixes are stacked in time delay order to form the tap coefficient matrix. Each symbol vector in the multi-load symbol sequence is concatenated with several adjacent symbol vectors to form an extended symbol vector. The extended symbol vector is then multiplied by the tap coefficient matrix to obtain the filtered output vector. Arrange the filtered output vectors corresponding to all symbol vectors in chronological order to obtain the parallel symbol stream after interference suppression.

7. The adaptive interference suppression method for multi-load parallel communication according to claim 6, characterized in that, The residual interference energy is calculated on the parallel symbol stream after interference suppression to generate a residual interference energy value. This residual interference energy value is then fed back to the recursive update step of the improved minimum mean square error algorithm to correct the estimated interference covariance matrix value for the next time step. This includes: Extract the filtered output value of the pilot symbol position from the parallel symbol stream after interference suppression, obtain the real value of the transmitted pilot symbol at the corresponding position, and perform a symbol-by-symbol subtraction operation between the filtered output value and the real value of the transmitted pilot symbol to generate a residual symbol sequence. For each residual symbol in the residual symbol sequence, the magnitude squared is calculated, and the sum of the magnitude squared calculation results is divided by the number of residual symbols to obtain the residual interference energy value. The residual interference energy value is compared with a preset interference energy threshold value. When the residual interference energy value is greater than the interference energy threshold value, an enhancement update trigger flag is generated. Based on the enhanced update trigger flag, the forgetting factor of the recursive least squares algorithm is adjusted to an enhanced forgetting factor value that is less than the current value, and the adjusted forgetting factor value is used for the next time step weight calculation in the exponential weighted moving average processing corresponding to the interference covariance matrix estimate.

8. The adaptive interference suppression method for multi-load parallel communication according to claim 4, characterized in that, The method further includes adaptive step size adjustment processing for the recursive update step of the improved minimum mean square error algorithm, wherein the adaptive step size adjustment processing includes: Extract the error sign between the input symbol vector at the current time and the updated weight vector at the current time, and calculate the moving average magnitude of the error sign; The time-varying severity level of the current channel is determined based on the moving average amplitude value, and the time-varying severity level includes slow time-varying level, medium time-varying level and fast time-varying level; When the time-varying intensity is classified as slow time-varying, the forgetting factor of the recursive least squares algorithm is set to a large value within the first numerical range. When the time-varying intensity is classified as medium time-varying, the forgetting factor is set to a medium value within the second numerical range. When the time-varying intensity is classified as fast time-varying, the forgetting factor is set to a small value within the third numerical range. The set forgetting factor is used as the control parameter for the current iteration step size of the recursive least squares algorithm, and is used to calculate the updated values ​​of the gain vector and the weight vector at the current time.

9. The adaptive interference suppression method for multi-load parallel communication according to claim 4, characterized in that, The method further includes a step of sparsifying the interference coefficient matrix, wherein the sparsification process includes: Iterate through all off-diagonal elements in the interference coefficient matrix at the current time and calculate the magnitude of each off-diagonal element; The modulus of each off-diagonal element is compared with a preset sparsification threshold. When the modulus of the off-diagonal element is less than the sparsification threshold, the value of the corresponding off-diagonal element is set to zero. When the modulus of the off-diagonal element is greater than or equal to the sparsification threshold, the original value of the corresponding off-diagonal element is retained. The sparsity of the matrix is ​​obtained by calculating the proportion of the total number of non-zero elements in the interference coefficient matrix after zeroing out the total number of elements. When the sparsity of the matrix is ​​less than the preset lower limit threshold of sparsity, the sparsification threshold value is gradually reduced and the step of comparing the modulus value of each off-diagonal element with the preset sparsification threshold value is repeated until the sparsity of the matrix is ​​greater than or equal to the lower limit threshold of sparsity, and a sparsification interference coefficient matrix is ​​obtained. The sparsed interference coefficient matrix is ​​used to replace the original interference coefficient matrix in subsequent matrix multiplication operations.

10. The adaptive interference suppression method for multi-load parallel communication according to claim 1, characterized in that, The method further includes performing iterative interference cancellation processing on the interference-suppressed parallel symbol stream, the iterative interference cancellation processing including: The parallel symbol stream after interference suppression is used as the current iteration input symbol stream. Hard decision processing is performed on the current iteration input symbol stream to generate a decision symbol sequence. The decision symbol sequence is convolved with the estimated interference covariance matrix to generate a reconstructed interference signal sequence. Subtract the reconstructed interference signal sequence from the multi-load symbol sequence to generate the interference-cancelled residual symbol stream; The residual symbol stream after interference cancellation is used as the new current iteration input symbol stream. Hard decision processing, convolution operation on the decision symbol sequence, and subtraction operation on the multi-load symbol sequence are repeatedly performed on the current iteration input symbol stream until a preset number of iterations is reached or the energy of the reconstructed interference signal sequence is lower than a preset energy threshold. The final interference-cancelled symbol stream is then output as the enhanced interference suppression result.

Images