A joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration

Abstract

This invention discloses a joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration, belonging to the field of blind signal processing in communication. This method addresses the problems of error propagation caused by hard-decision regeneration and excessive computational complexity of soft-information regeneration in iterative blind receivers by introducing an adaptive hybrid decision regeneration mechanism into the iterative loop of channel estimation and data detection. This mechanism dynamically sets a threshold based on the confidence level of the posterior log-likelihood ratio output by the decoder, using hard decision for high-confidence information and soft decision for low-confidence information, and adaptively adjusting the threshold based on the target soft decision rate to balance performance and computational overhead. This invention can effectively suppress error propagation under conditions of unknown modulation coding and channel state, achieving reliable joint blind estimation and identification with controllable computational complexity.

Description

Technical Field

[0001] This invention belongs to the field of blind processing technology for communication signals, and particularly relates to a joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration. Background Technology

[0002] In modern wireless communication systems, critical applications such as cognitive radio, dynamic spectrum sharing, electronic reconnaissance, and non-cooperative communication place extremely high demands on blind signal processing techniques. In these scenarios, the receiver often has no knowledge of key parameters such as the modulation scheme, channel coding scheme, channel state information, and noise power of the received signal, yet it must reliably recover the original information under these conditions. Traditional blind processing methods typically employ a cascaded structure, separating and sequentially executing tasks such as channel estimation, modulation identification, and decoding. This simple approach has an inherent flaw: estimation errors generated in previous processing stages accumulate and propagate to subsequent stages, leading to a significant degrade in overall system performance.

[0003] To overcome the error accumulation problem, the joint processing approach has emerged. Recent research has proposed an iterative receiver framework, the core of which is to establish an iterative loop between the channel estimation and data detection modules, leveraging the ping-pong effect to mutually reinforce each other: more accurate channel estimation leads to more reliable data detection, while more reliable data detection results can be used as pseudo-pilots to improve the accuracy of channel estimation. Theoretically, this framework can approach optimal performance, becoming a cutting-edge direction in blind signal processing. However, in-depth analysis of the internal mechanism of this iterative framework reveals a crucial aspect that has been generally overlooked in existing research: its information regeneration mechanism. Currently, the mainstream approach for this mechanism is hard-decision regeneration.

[0004] The fatal weakness of hard-decision regeneration mechanisms lies in their extreme sensitivity to the confidence level of the decoder output. Even with strong error correction coding, the log-likelihood ratio confidence level of the decoder output is still insufficient under low signal-to-noise ratio conditions. In such cases, hard-decision processing introduces non-negligible errors. These erroneous symbols severely contaminate the regenerated pseudo-pilot sequence, leading to biases in the next round of channel estimation and further deteriorating subsequent decoding performance. Ultimately, this traps the entire iterative process in a vicious cycle of performance degradation, severely limiting the system's performance floor under adverse channel conditions.

[0005] As an upper bound for performance comparison, the soft information regeneration scheme retains probabilistic information losslessly by calculating the expected value of symbols, theoretically avoiding the aforementioned error propagation. However, this scheme requires calculating the expected value across the entire modulation constellation diagram, and its computational complexity increases exponentially with the modulation order, making it difficult to implement in engineering practice, especially for applications with high real-time requirements. Therefore, there is an urgent need in this field for an innovative information regeneration mechanism that can effectively overcome the error propagation bottleneck caused by hard-decision regeneration, while avoiding the unbearable computational overhead introduced by soft information regeneration, thus achieving a practical and intelligent balance between system performance and implementation complexity. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration, thereby resolving the issues present in the prior art.

[0007] In a first aspect, to achieve the above objectives, the present invention provides a joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration, wherein the receiver processes the received signal under the condition that the modulation and coding scheme is unknown, and the method includes:

[0008] Iterate through the predefined set of candidate modulation and coding schemes, and for each candidate scheme as the current hypothesis, execute the inner iteration process;

[0009] Perform decoder-assisted expectation maximization initialization for the current assumption to obtain initial estimates of the channel parameters;

[0010] Based on the initial estimate, an iterative loop is performed between the channel estimation module and the data detection and regeneration module;

[0011] In each inner iteration, the channel parameter estimate is updated using the pseudo-pilot symbol sequence regenerated in the previous round through the expectation-maximization algorithm.

[0012] The received signal is equalized and soft demodulated using the updated channel parameter estimation. After soft input and soft output decoding, a new round of pseudo pilot symbol sequence is generated through an adaptive hybrid decision regeneration mechanism.

[0013] Determine whether the inner iteration satisfies the convergence condition or reaches the maximum number of iterations. If it does, stop the inner iteration and calculate the log-likelihood value of the current hypothesis.

[0014] After all candidate schemes have completed their inner-layer iterations, the log-likelihood values ​​of each hypothesis are compared. The candidate scheme with the maximum value is taken as the final modulation and coding recognition result, and its corresponding decoded information bits are output.

[0015] Optionally, the decoder-assisted expectation maximization initialization process includes:

[0016] The channel parameters are roughly estimated using the higher-order cumulants of the received signal to obtain a rough estimate.

[0017] Starting from the rough estimate, perform several lightweight expectation-maximization pre-iterations;

[0018] In each pre-iteration, the symbol posterior probability and the soft symbol expectation are calculated, and the soft symbol expectation is used as the pseudo-pilot to update the channel parameters;

[0019] The channel parameter values ​​obtained after the pre-iteration are used as the initial estimates for the inner iterative process.

[0020] Optionally, the process of updating the channel parameter estimate using the expectation-maximization algorithm includes:

[0021] In the expectation step, the conditional expectation of each path signal component in the multipath channel is calculated;

[0022] In the maximization step, based on the conditional expectation, the channel coefficients and noise power of each path are independently updated by solving the least squares problem.

[0023] Optionally, the process by which the adaptive hybrid decision regeneration mechanism generates a new round of pseudo-pilot symbol sequences includes:

[0024] The decision threshold is dynamically determined based on the distribution of the absolute value of the posterior log-likelihood ratio output by the soft-input soft-output decoder and the preset target soft decision rate.

[0025] Based on the aforementioned decision threshold, the coded bits constituting each modulation symbol are divided into a high-confidence set and a low-confidence set;

[0026] Hard decision is made on the bits in the high confidence set, and all consistent symbol points are selected from the complete constellation graph based on the hard decision results to form a consistent symbol subset.

[0027] Based on the posterior probability of the bits in the low-confidence set, the conditional expectation of the symbols is calculated on the subset of consistent symbols to generate the new round of pseudo-pilot symbols.

[0028] Optionally, the process of dynamically determining the decision threshold includes:

[0029] Collect the absolute values ​​of the posterior log-likelihood ratios output by the soft-input soft-output decoder;

[0030] Calculate the absolute value of each of the posterior log-likelihood ratios to form a set of absolute values;

[0031] Calculate the quantiles in the absolute value set corresponding to the target soft decision rate, and set the quantiles as the decision threshold.

[0032] Optionally, the process of determining whether the inner iteration satisfies the convergence condition includes:

[0033] Calculate the mean square error between the channel parameter estimates obtained in the current iteration and the previous iteration;

[0034] If the mean square error is less than a preset first threshold, or the number of iterations reaches a preset second threshold, then convergence is determined.

[0035] Secondly, the present invention also provides a joint blind channel estimation and modulation identification system based on adaptive hybrid regeneration, for implementing a joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration, the system comprising:

[0036] The assumption traversal module is used to traverse a predefined set of candidate modulation and coding schemes when the modulation and coding schemes are unknown, and output each candidate scheme as the current hypothesis.

[0037] An initialization module is used to perform decoder-assisted expectation maximization initialization for the current assumption to obtain initial estimates of the channel parameters;

[0038] An iterative control module is used to initiate and control an iterative loop between the channel estimation module and the data detection and regeneration module based on the initial estimate.

[0039] The channel estimation module is used in each iteration to update the channel parameter estimation by using the pseudo-pilot symbol sequence output by the data detection and regeneration module in the previous round and through the expectation-maximization algorithm.

[0040] The data detection and regeneration module is used to perform equalization and soft demodulation on the received signal using the updated channel parameter estimation, and then generate a new round of pseudo-pilot symbol sequence through an adaptive hybrid decision regeneration mechanism after soft input and soft output decoding.

[0041] The convergence judgment and likelihood calculation module is used to determine whether the iteration for the current hypothesis meets the convergence condition or reaches the maximum number of iterations, and stops the iteration when it meets the condition, and calculates the log-likelihood value of the current hypothesis.

[0042] The final decision module is used to compare the log-likelihood values ​​of each hypothesis after all candidate schemes have completed iterations, take the candidate scheme with the maximum value as the final modulation and coding recognition result, and output its corresponding decoded information bits.

[0043] Thirdly, the present invention also provides a computer terminal device, comprising:

[0044] One or more processors;

[0045] A memory, coupled to the processor, for storing one or more programs;

[0046] When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration in the first aspect described above.

[0047] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration described in the first aspect above.

[0048] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration described in the first aspect.

[0049] Compared with the prior art, the present invention has the following advantages and technical effects:

[0050] This invention provides a joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration. By introducing a confidence-based adaptive hybrid decision regeneration mechanism into the iterative blind processing framework, this invention effectively suppresses the performance degradation cycle caused by hard-decision error propagation under low signal-to-noise ratio conditions, achieving performance close to the soft information regeneration benchmark requiring full constellation diagram calculation. Simultaneously, this invention achieves flexible control of computational complexity and intelligent trade-off between computational complexity and performance gain by selectively performing soft decisions on low-confidence information and dynamically adjusting the decision threshold according to a preset target soft decision rate. This adaptive strategy enables the algorithm to adapt to different channel conditions, improving the overall robustness of the system. Ultimately, this invention solves the key error propagation problem in iterative blind receivers with significantly lower computational overhead than traditional soft-decision schemes, possessing both good performance and practical engineering value. Attached Figure Description

[0051] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0052] Figure 1 This is a flowchart illustrating the overall architecture of the iterative joint blind estimation and identification method according to an embodiment of the present invention.

[0053] Figure 2 This is a flowchart of the soft-input soft-output data detection and regeneration module according to an embodiment of the present invention;

[0054] Figure 3This is a flowchart of the adaptive hybrid decision regeneration module algorithm according to an embodiment of the present invention;

[0055] Figure 4 This is a schematic diagram of adaptive decision threshold selection based on LLR distribution according to an embodiment of the present invention;

[0056] Figure 5 This is a flowchart of the decoder-assisted expectation maximization initialization method according to an embodiment of the present invention. Detailed Implementation

[0057] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0058] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0059] Example 1

[0060] To overcome the shortcomings of the hard decision regeneration (HDR) mechanism in existing iterative blind processing frameworks, which introduces error propagation at low signal-to-noise ratios, and the problem of excessive computational complexity of the soft information regeneration (SIR) mechanism, this invention proposes an iterative joint blind estimation and recognition method based on adaptive hybrid decision regeneration (A-HyDR).

[0061] The core idea of ​​this method is to no longer apply hard or soft decision to all decoded information in the same way. Instead, it uses the log-likelihood ratio (LLR) output by the decoder as a confidence measure: hard decision is used for information with high confidence (large absolute value of LLR), which is simple to calculate; while soft decision is used only for ambiguous information with low confidence (small absolute value of LLR).

[0062] More importantly, this invention proposes an adaptive threshold strategy based on the LLR distribution, which sets a target "computational cost" (i.e., the target soft decision rate). To dynamically deduce the judgment threshold. This allows the algorithm to intelligently balance performance gains and computational costs, automatically adapting to different channel conditions.

[0063] like Figure 1 As shown, this embodiment provides a joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration. The receiver processes the received signal under conditions where the modulation and coding schemes are unknown. The method includes:

[0064] Iterate through the predefined set of candidate modulation and coding schemes, and for each candidate scheme as the current hypothesis, execute the inner iteration process;

[0065] Perform decoder-assisted expectation maximization initialization for the current assumption to obtain initial estimates of the channel parameters;

[0066] Based on the initial estimate, an iterative loop is performed between the channel estimation module and the data detection and regeneration module;

[0067] In each inner iteration, the channel parameter estimate is updated using the pseudo-pilot symbol sequence regenerated in the previous round through the expectation-maximization algorithm.

[0068] The received signal is equalized and soft demodulated using the updated channel parameter estimation. After soft input and soft output decoding, a new round of pseudo pilot symbol sequence is generated through an adaptive hybrid decision regeneration mechanism.

[0069] Determine whether the inner iteration satisfies the convergence condition or reaches the maximum number of iterations. If it does, stop the inner iteration and calculate the log-likelihood value of the current hypothesis.

[0070] After all candidate schemes have completed their inner-layer iterations, the log-likelihood values ​​of each hypothesis are compared. The candidate scheme with the maximum value is taken as the final modulation and coding recognition result, and its corresponding decoded information bits are output.

[0071] As one implementation method in this embodiment, the decoder-assisted expectation maximization initialization process includes:

[0072] The channel parameters are roughly estimated using the higher-order cumulants of the received signal to obtain a rough estimate.

[0073] Starting from the rough estimate, perform several lightweight expectation-maximization pre-iterations;

[0074] In each pre-iteration, the symbol posterior probability and the soft symbol expectation are calculated, and the soft symbol expectation is used as the pseudo-pilot to update the channel parameters;

[0075] The channel parameter values ​​obtained after the pre-iteration are used as the initial estimates for the inner iterative process.

[0076] Specifically, it includes:

[0077] S1: Initialization:

[0078] Traverse a predefined set of candidate modulation and coding schemes ,in This is the modulation candidate set for the transmitter. Let be the channel coding candidate set. Assume the unknown modulation and coding scheme (MCS) is . ,in Indicates the modulation format of the transmitted signal. This represents the channel coding scheme for the transmitted signal. For each candidate scheme in the set... This is treated as the current hypothesis, and the complete inner iteration process from S2 to S7 is executed for this hypothesis. Before the inner loop begins, the current hypothesis is... Perform a decoder-assisted expectation-maximization (DA-EM) initialization to provide a robust initial estimate of the channel parameters for the main iterative engine (S3-S6). ,here This is the estimated parameter vector. The channel parameter vector... It contains all the unknown channel scalar parameters to be estimated. Specifically, for a given channel with... Multipath channel with path 1, the 2nd path Complex channel coefficients of each path It can be derived from its real part and imaginary part, or equivalently from its amplitude. and phase To characterize (i.e.) ),here Therefore, the complete vector of parameters to be estimated. That is, the magnitude of all paths. Phase and noise power The set of parameters can be represented as:

[0079] ;

[0080] Furthermore, the decoder-assisted expectation-maximization (DA-EM) initialization method in step S1 specifically includes:

[0081] S1.1: A rough estimate based on higher-order cumulants: utilizing the first... Received signal at time 1 The higher-order cumulants are calculated. Since the higher-order cumulants of Gaussian noise are zero, this method can naturally eliminate the influence of noise and construct the channel coefficients for each path. Solving the nonlinear equations yields a rough initial estimate of the channel parameters. .

[0082] S1.2: Lightweight EM fine-tuning: using a coarse initial estimate of channel parameters Starting from this point, proceed Five lightweight EM iterations (here, five iterations) are performed. In each iteration:

[0083] Lightweight E-Step: Calculates each time step using only the channel model, without using the SISO channel decoder. The sign posterior probability .here For the first Transmission symbols of time, Indicating the first in the modulation constellation diagram Each constellation point, For the first The received symbol at that moment, Indicates the first The parameter estimates for the second time.

[0084] Lightweight M-Step: Calculate the first... The expected value of the soft sign of the second time Using this soft symbol sequence as a pseudo-pilot, and substituting it into the EM-CNE update formula in S3, a more accurate result is obtained. Secondary channel parameter estimation .

[0085] go through After the iteration, the channel parameter estimates are obtained. As the initial value of the channel parameters of S2 .

[0086] As one implementation method in this embodiment, the process of updating the channel parameter estimate using the expectation-maximization algorithm includes:

[0087] In the expectation step, the conditional expectation of each path signal component in the multipath channel is calculated;

[0088] In the maximization step, based on the conditional expectation, the channel coefficients and noise power of each path are independently updated by solving the least squares problem.

[0089] Specifically, it includes:

[0090] S3: Expectation-Maximization Channel Estimation (EM-CNE): In the... In the iteration, the SISO module is used in the ... The pseudo-pilot symbol sequence output by the next iteration and received signals The multipath channel parameter vector is updated using the Expectation-Maximization (EM) algorithm (including E-Step and M-Step). The estimation yields the final channel parameter estimation vector. .

[0091] As one implementation method in this embodiment, the process of the adaptive hybrid decision regeneration mechanism generating a new round of pseudo-pilot symbol sequences includes:

[0092] The decision threshold is dynamically determined based on the distribution of the absolute value of the posterior log-likelihood ratio output by the soft-input soft-output decoder and the preset target soft decision rate.

[0093] Based on the aforementioned decision threshold, the coded bits constituting each modulation symbol are divided into a high-confidence set and a low-confidence set;

[0094] Hard decision is made on the bits in the high confidence set, and all consistent symbol points are selected from the complete constellation graph based on the hard decision results to form a consistent symbol subset.

[0095] Based on the posterior probability of the bits in the low-confidence set, the conditional expectation of the symbols is calculated on the subset of consistent symbols to generate the new round of pseudo-pilot symbols.

[0096] S4: Soft-Input Soft-Output (SISO) Data Detection and Regeneration: Utilizing the EM-CNE module in the first... Channel parameters updated in the next iteration For the received signal After processing, a new, more reliable first-generation [processor] will be generated. Secondary pseudo-pilot symbol sequence Step S4 specifically includes:

[0097] S4.1: Equalization and Soft Demodulation: Utilizing the first Secondary channel estimation And the pseudo pilot generated in the previous round (Used to eliminate inter-symbol interference (ISI)) for the received signal Equalization is performed, the posterior probability of each modulation symbol is calculated, and it is converted into the channel LLR, i.e., the channel log-likelihood ratio, for each coded bit.

[0098] S4.2: SISO Channel Decoding: Input the channel log-likelihood ratio (LLR) obtained in S5.1 into the current hypothesis. The corresponding SISO decoder, utilizing codeword constraints, outputs a significantly more reliable coded bit a posteriori (LLR) output. That is, the posterior log-likelihood ratio.

[0099] S4.3: Adaptive Hybrid Decision Regeneration (A-HyDR): This converts the a posteriori LLR output from S5.2 into a new LLR. Input the A-HyDR module proposed in this invention to generate a new round of pseudo-pilot symbol sequences. .

[0100] As one implementation method in this embodiment, the process of dynamically determining the decision threshold includes:

[0101] Collect the absolute values ​​of the posterior log-likelihood ratios output by the soft-input soft-output decoder;

[0102] Calculate the absolute value of each of the posterior log-likelihood ratios to form a set of absolute values;

[0103] Calculate the quantiles in the absolute value set corresponding to the target soft decision rate, and set the quantiles as the decision threshold.

[0104] Furthermore, the core adaptive hybrid decision regeneration (A-HyDR) method of this invention in step S4.3 specifically includes:

[0105] S4.3.1: Determine the adaptive threshold :

[0106] a. Set a target soft decision rate This value represents an acceptable computational complexity (e.g., =15%, indicating that only the most uncertain 15% of bits are willing to be soft-determined.

[0107] b. Collect the posterior LLR of all coded bits in the current data block output by S4.2, and calculate the absolute value of their log-likelihood ratios. .

[0108] c. Find these log-likelihood ratios Target soft decision rate of absolute value set Quantity (i.e., the value that is ranked in ascending order at the 15th percentile).

[0109] d. Set this quantile as the decision threshold for the current data block. .

[0110] S4.3.2: Bit Blocking and Processing: For the components of the first... Modulation symbols The set of coded bits and their log-likelihood ratios LLR :

[0111] a. High Confidence Set (HCS): Find all sets that satisfy... The bits are used to make hard decisions on them. The value is set to 0. If the result is determined to be 1, a set of definite bit values ​​and their indices are obtained.

[0112] b. Low Confidence Set (LCS): Find all sets that satisfy... The bits are stored, their LLR values ​​are preserved, and the posterior probabilities of being 0 and 1 are calculated.

[0113] S4.3.3: Constructing a subset of consistent symbols That is, assuming a complete set of constellation diagrams corresponding to the current modulation method:

[0114] Traversing the current modulation scheme assumptions Complete collection of constellation charts Select all symbols that do not conflict with the HCS hard decision result in S4.3.2 to form a consistent symbol subset. .

[0115] S4.3.4: Calculate the conditional expectation (soft sign):

[0116] a. For subsets Each symbol in Its conditional probability is determined solely by the bit probabilities in the LCS:

[0117] ;

[0118] in, For the first One transmission symbol, For the received signal vector, For the first in the consistent symbol subset Each constellation point, For the first The first symbol corresponds to the first One encoded bit, Let be the bit matching probability, and LCS be the set of consistent matches.

[0119] b. Regarding the above in the subset Normalize all the probabilities calculated above so that their sum is 1.

[0120] c. Calculate the normalized conditional expectation value to obtain the first... A hybrid regeneration symbol :

[0121] ;

[0122] in, For the first in the consistent symbol subset Each constellation point, This represents the normalized conditional probability. Indicates the first One transmission symbol, Here, denoted as the received signal vector, and HCS represents the previously obtained hard decision result.

[0123] S4.3.5: Output: All indivual Combined into a pseudo-pilot symbol sequence And output it for S3 in the next iteration. Use when needed.

[0124] As one implementation method in this embodiment, the process of determining whether the inner iteration satisfies the convergence condition includes:

[0125] Calculate the mean square error between the channel parameter estimates obtained in the current iteration and the previous iteration;

[0126] If the mean square error is less than a preset first threshold, or the number of iterations reaches a preset second threshold, then convergence is determined.

[0127] S5: Inner Loop Convergence Decision: Checks whether the inner loop meets the convergence criteria. Calculates the current channel parameter estimate. Compared with the previous estimate The mean square error (MSE) between them. If ( For the threshold, Value )or (here If the loop converges to 50, the inner loop is considered to have converged, the iteration stops, and the process proceeds to S7. If it has not converged, then... Return to S4 and continue to the next iteration.

[0128] S6: Likelihood calculation: For the current hypothesis Once its inner loop (S3-S6) converges, the final channel estimate is used. and pseudo pilot Calculate the observed received signal under this assumption. Total log-likelihood .

[0129] S7: Final decision: After traversing all candidate solutions in the outer loop (S1) Next, compare the list of log-likelihood values ​​for all hypotheses. Select the hypothesis that maximizes the log-likelihood value. As the final identification result:

[0130] ;

[0131] in, The total log-likelihood value is selected to ensure that the "received signal" is... Total log-likelihood value under this candidate scheme "The candidate solution that reaches the maximum value" .

[0132] And in The final information bit estimate obtained under the assumptions As the decoded output.

[0133] This embodiment uses a blind receiver employing LDPC codes and operating in a multipath frequency-selective fading channel as an example to illustrate the method of the present invention in detail.

[0134] Reference Figure 1 An iterative joint blind estimation and identification method based on adaptive hybrid regeneration includes the following steps:

[0135] S1: Initialization:

[0136] like Figure 1 As shown, this invention employs a two-level nested iterative algorithm architecture. The outer loop is a hypothesis testing process based on maximum likelihood.

[0137] When the receiver starts working, it only knows the sequence of received signals. and a set of candidate modulation and coding schemes This set represents all possible modulation schemes. (e.g., QPSK, 8PSK, 16-QAM) and encoding schemes (e.g., the Cartesian product of LDPCR=1 / 2, LDPCR=2 / 3).

[0138] Step S1 is to traverse the set. Each specific candidate solution ,in This refers to the candidate modulation format corresponding to the candidate scheme. This refers to the candidate channel coding scheme corresponding to the candidate scheme. For each hypothetical scheme... The algorithm "assumes" that it is correct and performs a complete inner iteration process from S2 to S7 for it.

[0139] S2: Initialization (DA-EM scheme):

[0140] The performance of any iterative algorithm is highly dependent on the choice of its initial values. A poor starting point can lead to slow convergence or getting trapped in local optima. Therefore, before the inner loop (S3) begins, this invention employs an efficient decoder-assisted expectation-maximization (DA-EM) initialization scheme to provide a robust starting point for the channel parameters of the main iterative engine. .

[0141] Reference Figure 5 The initialization scheme (DA-EM) specifically includes two phases:

[0142] Phase 1: Rough estimation based on higher-order cumulants.

[0143] This stage utilizes received signals. The statistical properties of the Gaussian random variable (noise) are as follows: the cumulants of third order and above are strictly zero. Therefore, by calculating the received signal... The higher-order cumulants naturally eliminate the influence of noise. By solving the cumulant equations, the channel parameters (i.e., amplitude) can be obtained. Phase A rough estimate of (etc.) .

[0144] Phase Two: Lightweight EM Fine-tuning.

[0145] The goal of this stage is to quickly refine the coarse values ​​before entering the main loop (S4.3) which contains the complex channel decoder.

[0146] This phase of execution The third (3-5th) iteration. In the next pre-iteration:

[0147] Lightweight E-Step: Based on the previous channel estimation parameter values Calculate each time step The sign posterior probability Note: This step does not involve any channel decoding; it only utilizes the channel model, resulting in minimal computational overhead.

[0148] Lightweight M-Step: Computing the Expectation of Soft Symbols Then, using this high-quality soft symbol sequence... As a "pseudo-pilot," it is substituted into the EM-CNE update formula in S3 to obtain a more accurate channel parameter estimate. .

[0149] go through The final channel parameters obtained after the iteration This will be used as the initial value for the S2 main iteration engine. .

[0150] S2: Inner loop (iterative estimation and detection):

[0151] like Figure 1 As shown, for a given hypothesis and initialization parameters The inner loop aims to iteratively optimize channel estimation and data detection using the "ping-pong effect." An iteration counter is set. .

[0152] S3: Expectation-Maximization Channel Estimation (EM-CNE):

[0153] The goal of this module is: in the... In this iteration, S4 is used in the previous round ( The pseudo-pilot symbol sequence provided To update the multipath channel parameter vector defined in the "Summary of the Invention" section. The estimate.

[0154] In multipath channels, due to the presence of inter-symbol interference (ISI), directly solving the ML estimation problem is a challenging non-convex optimization problem. This invention employs the EM algorithm for iterative solution.

[0155] E-Step (Expectation Step): Computing the unobservable "complete data". (i.e., the first) Conditional expectation of the signal components contributed by each path .

[0156] M-Step (Maximization Step): Using the calculated expected value For each path, a simple least squares problem is solved independently to obtain the channel parameters (i.e. , and The closed-form update solution of ).

[0157] The updated formula is as follows:

[0158] Channel coefficient update ( ):

[0159] ;

[0160] in For the first One regenerated modulation symbol, For the first Intermediate values ​​in the next iteration. For the first During the nth iteration, the 1st Path number An estimate of a complete dataset. For the first During the nth iteration, the 1st One regenerated modulation symbol.

[0161] Noise power update ( ):

[0162] ;

[0163] in It is the total number of received symbols. For the first One received symbol, For the first Channel coefficient of the second iteration × the first iteration Delayed regeneration symbol for the next iteration.

[0164] After S3, the current number is obtained. The latest channel estimate in the next iteration .

[0165] S4: SISO Data Detection and Regeneration:

[0166] The goal of this module is to utilize the channel parameters that S3 just updated. For the received signal After processing, a new round of more reliable pseudo-pilot symbols is finally output. .

[0167] Reference Figure 2 The SISO module contains three cascaded sub-steps:

[0168] S4.1: Equalization and Soft Demodulation:

[0169] This step involves simulating the received signal. This is converted into bit soft information (channel LLR) that the SISO decoder can understand.

[0170] S4.1.1: Equalization: In multipath channels, the received signal... Not only with the current symbol Related to, and also related to the previous The symbol (ISI) is related. Defined under the assumption... The expected signal (including the ISI estimate) at time is :

[0171] ;

[0172] in It comes from the previous round pseudo-pilot, This is the channel coefficient estimate for the 0th path (in a multipath channel, the 0th path corresponds to the channel attenuation / phase of the "current symbol"). For the first Channel coefficient estimates for each path.

[0173] The symbolic posterior probability is calculated based on the Gaussian noise model and Bayes' theorem. for:

[0174] ;

[0175] in, For the estimated noise power, The squared error between the received signal and the desired signal is given.

[0176] S4.1.2: Soft demodulation: utilizing By accumulating the probabilities of the corresponding symbol sets, the coded bit is calculated. The LLR channel.

[0177] S4.2: SISO channel decoding:

[0178] This module is responsible for receiving the channel LLR output from S4.1 and utilizing codeword constraints (based on the current assumptions). The a posteriori LLR (i.e., LDPC code) output reliability is significantly enhanced. .

[0179] S4.3: Adaptive Hybrid Decision Regeneration (A-HyDR):

[0180] This is the final link in the iterative loop, and also the core innovation of this invention. (Refer to...) Figure 3 The A-HyDR scheme proposed in this invention has the following process:

[0181] S4.3.1: Determine the adaptive threshold :

[0182] Reference Figure 4 The threshold of the present invention It is dynamically adaptive. It is not fixed. Instead, it fixes a computational cost that is willing to be paid, namely the target soft decision rate. (You can set it to 15% first).

[0183] The algorithm is as follows:

[0184] a. Collect the absolute values ​​of all LLRs in the current data block output by S4.2. .

[0185] b. Find these absolute values Quantiles. For example, if =15%, then find all The value that ranks in the 15th percentile from smallest to largest.

[0186] c. Set this quantile as the decision threshold for the current data block. .

[0187] S4.3.2: Bit blocks:

[0188] For constituting the first a symbol LLR :

[0189] High confidence set (HCS): All sets that satisfy... 0 bits.

[0190] Low Confidence Set (LCS): All sets that satisfy... 0 bits.

[0191] S4.3.3: Constructing a subset of consistent symbols :

[0192] Perform hard decisions on the bits in the HCS. Then, from the complete constellation diagram... In the process, all symbol points that are consistent with the HCS hard decision result are selected to form a subset. .

[0193] S4.3.4: Calculation of expected conditions:

[0194] Only in subsets Above, soft decision is performed using the probability of LCS, and soft symbol calculation is performed. .

[0195] S4.3.5: Output:

[0196] Output As This is used for the next iteration of S4.

[0197] S5: Inner loop convergence decision:

[0198] Check the stopping condition of the inner loop.

[0199] Parameter stability criteria: (Pick ).

[0200] Maximum number of iterations criterion: (Pick ).

[0201] If any one of the conditions is met, the inner loop stops and proceeds to S6. Otherwise, And return to S2.

[0202] S6: Likelihood value calculation:

[0203] When the inner loop is based on the assumption After convergence, the final parameter estimates are used. and regeneration symbols Calculate the total log-likelihood function value under this assumption. .

[0204] S7: Final Verdict:

[0205] When the outer loop of S1 has traversed all... Afterwards, a list of likelihood scores will be obtained. Following the maximum likelihood (ML) criterion, select the scores that make the likelihood scores the most likely. The hypothesis with the largest value As the final identification result.

[0206] The effectiveness of this mechanism lies in the fact that incorrect assumptions... This can lead to model mismatch, causing the inner loop to fail to converge correctly, thus systematically generating a low likelihood value, making it possible for the correct solution to be identified with a high probability.

[0207] Based on this, the present invention provides a joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration, which achieves performance close to high-complexity benchmarks while realizing better computational efficiency and controllability through hybrid decision-making. Its adaptive threshold strategy helps improve the robustness of the algorithm under different channel conditions, demonstrating potential application value in solving the error propagation problem in iterative blind receivers.

[0208] Example 2

[0209] In this embodiment, a computer terminal device is provided, including:

[0210] One or more processors;

[0211] A memory, coupled to the processor, for storing one or more programs;

[0212] When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the above-described joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration.

[0213] In this embodiment, a computer-readable storage medium is also provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the above-described joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration.

[0214] In this embodiment, an electronic device is also provided, including a memory and a processor. The memory stores a computer program, and the processor is configured to run the computer program to perform the steps of the above-described joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration.

[0215] In this embodiment, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the above-described joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration.

[0216] The aforementioned program can run on a processor or be stored in memory (or a computer-readable medium). Computer-readable media includes both permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0217] These computer programs may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes can be implemented by different modules for different steps.

[0218] This embodiment provides such an apparatus or system. The system, referred to as a joint blind channel estimation and modulation identification system based on adaptive hybrid regeneration, includes:

[0219] The assumption traversal module is used to traverse a predefined set of candidate modulation and coding schemes when the modulation and coding schemes are unknown, and output each candidate scheme as the current hypothesis.

[0220] An initialization module is used to perform decoder-assisted expectation maximization initialization for the current assumption to obtain initial estimates of the channel parameters;

[0221] An iterative control module is used to initiate and control an iterative loop between the channel estimation module and the data detection and regeneration module based on the initial estimate.

[0222] The channel estimation module is used in each iteration to update the channel parameter estimation by using the pseudo-pilot symbol sequence output by the data detection and regeneration module in the previous round and through the expectation-maximization algorithm.

[0223] The data detection and regeneration module is used to perform equalization and soft demodulation on the received signal using the updated channel parameter estimation, and then generate a new round of pseudo-pilot symbol sequence through an adaptive hybrid decision regeneration mechanism after soft input and soft output decoding.

[0224] The convergence judgment and likelihood calculation module is used to determine whether the iteration for the current hypothesis meets the convergence condition or reaches the maximum number of iterations, and stops the iteration when it meets the condition, and calculates the log-likelihood value of the current hypothesis.

[0225] The final decision module is used to compare the log-likelihood values ​​of each hypothesis after all candidate schemes have completed iterations, take the candidate scheme with the maximum value as the final modulation and coding recognition result, and output its corresponding decoded information bits.

[0226] As one implementation method in this embodiment, the initialization module includes:

[0227] The coarse estimation unit is used to perform a coarse estimation of the channel parameters using the higher-order cumulants of the received signal, and obtain a coarse estimate value.

[0228] A pre-iteration unit is used to control the execution of several lightweight expectation maximization pre-iterations starting from the coarse estimate;

[0229] In each pre-iteration, the symbol posterior probability and the soft symbol expectation are calculated, and the soft symbol expectation is used as the pseudo-pilot to update the channel parameters. Finally, the channel parameter values ​​obtained after the pre-iteration are output as the initial estimate.

[0230] As one implementation method in this embodiment, the channel estimation module includes:

[0231] The expectation calculation unit is used to calculate the conditional expectation of each path signal component in the multipath channel during the expectation step.

[0232] The parameter update unit is used to independently update the channel coefficients and noise power of each path by solving the least squares problem based on the conditional expectation during the maximization step.

[0233] As one implementation method in this embodiment, the adaptive hybrid decision regeneration mechanism in the data detection and regeneration module includes:

[0234] The threshold determination unit is used to dynamically determine the decision threshold based on the distribution of the absolute value of the posterior log-likelihood ratio output by the soft-input soft-output decoder and the preset target soft decision rate.

[0235] A bit grouping unit is used to divide the coded bits constituting each modulation symbol into a high-confidence set and a low-confidence set according to the decision threshold.

[0236] The symbol subset construction unit is used to perform hard decision on the bits in the high confidence set, and based on the hard decision result, select all consistent symbol points from the complete constellation graph to form a consistent symbol subset;

[0237] A soft symbol generation unit is used to calculate the conditional expectation value of symbols on the consistent symbol subset based on the posterior probability of bits in the low-confidence set, so as to generate the new round of pseudo-pilot symbols.

[0238] As one implementation method in this embodiment, the threshold determination unit is specifically used for:

[0239] Collect the absolute values ​​of the posterior log-likelihood ratios output by the soft-input soft-output decoder;

[0240] Calculate the quantiles of the target soft decision rate in the set of absolute values ​​of the posterior log-likelihood ratios;

[0241] The quantile is set as the decision threshold.

[0242] As one implementation method in this embodiment, the convergence judgment function in the convergence judgment and likelihood calculation module is implemented in the following way:

[0243] Calculate the mean square error between the channel parameter estimates output by the channel estimation module in the current iteration and the previous iteration;

[0244] If the mean square error is less than a preset first threshold, or the number of iterations reaches a preset second threshold, then convergence is determined.

[0245] The system or apparatus is used to implement the functions of the methods in the above embodiments. Each module in the system or apparatus corresponds to each step in the method, as has been described in the method and will not be repeated here.

[0246] The above implementation method solves the problem of joint blind channel estimation and modulation identification based on adaptive hybrid regeneration in related technologies, thereby ensuring that the problems existing in the prior art are resolved.

[0247] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A joint blind channel estimation and modulation identification method based on adaptive hybrid regeneration, characterized in that, The receiver processes the received signal under conditions where the modulation and coding scheme is unknown. The method includes: Iterate through the predefined set of candidate modulation and coding schemes, and for each candidate scheme as the current hypothesis, execute the inner iteration process; Perform decoder-assisted expectation maximization initialization for the current assumption to obtain initial estimates of the channel parameters; Based on the initial estimate, an iterative loop is performed between the channel estimation module and the data detection and regeneration module; In each inner iteration, the channel parameter estimate is updated using the pseudo-pilot symbol sequence regenerated in the previous round through the expectation-maximization algorithm. The received signal is equalized and soft demodulated using the updated channel parameter estimation. After soft input and soft output decoding, a new round of pseudo pilot symbol sequence is generated through an adaptive hybrid decision regeneration mechanism. The process by which the adaptive hybrid decision regeneration mechanism generates a new round of pseudo-pilot symbol sequences includes: The decision threshold is dynamically determined based on the distribution of the absolute value of the posterior log-likelihood ratio output by the soft-input soft-output decoder and the preset target soft decision rate. Based on the aforementioned decision threshold, the coded bits constituting each modulation symbol are divided into a high-confidence set and a low-confidence set; Hard decision is made on the bits in the high confidence set, and all consistent symbol points are selected from the complete constellation graph based on the hard decision results to form a consistent symbol subset. Based on the posterior probability of the bits in the low-confidence set, the conditional expectation of the symbols is calculated on the subset of consistent symbols to generate the new round of pseudo-pilot symbols. The process of dynamically determining the decision threshold includes: Collect the absolute values ​​of the posterior log-likelihood ratios output by the soft-input soft-output decoder; Calculate the absolute value of each of the posterior log-likelihood ratios to form a set of absolute values; Calculate the quantiles in the absolute value set corresponding to the target soft decision rate, and set the quantiles as the decision threshold; Determine whether the inner iteration satisfies the convergence condition or reaches the maximum number of iterations. If it does, stop the inner iteration and calculate the log-likelihood value of the current hypothesis. After all candidate schemes have completed their inner-layer iterations, the log-likelihood values ​​of each hypothesis are compared. The candidate scheme with the maximum value is taken as the final modulation and coding recognition result, and its corresponding decoded information bits are output.

2. The method according to claim 1, characterized in that, The decoder-assisted expectation maximization initialization process includes: The channel parameters are roughly estimated using the higher-order cumulants of the received signal to obtain a rough estimate. Starting from the rough estimate, perform several lightweight expectation-maximization pre-iterations; In each pre-iteration, the symbol posterior probability and the soft symbol expectation are calculated, and the soft symbol expectation is used as the pseudo-pilot to update the channel parameters; The channel parameter values ​​obtained after the pre-iteration are used as the initial estimates for the inner iterative process.

3. The method according to claim 1, characterized in that, The process of updating the channel parameter estimate using the expectation-maximization algorithm includes: In the expectation step, the conditional expectation of each path signal component in the multipath channel is calculated; In the maximization step, based on the conditional expectation, the channel coefficients and noise power of each path are independently updated by solving the least squares problem.

4. The method according to claim 1, characterized in that, The process of determining whether the inner iteration satisfies the convergence condition includes: Calculate the mean square error between the channel parameter estimates obtained in the current iteration and the previous iteration; If the mean square error is less than a preset first threshold, or the number of iterations reaches a preset second threshold, then convergence is determined.

5. A joint blind channel estimation and modulation identification system based on adaptive hybrid regeneration, characterized in that, The system for implementing the method according to any one of claims 1-4 comprises: The assumption traversal and control module is used to traverse a predefined set of candidate modulation and coding schemes and trigger the inner processing flow with each candidate scheme as the current assumption. The initialization module is used to perform decoder-assisted expectation maximization initialization for the current hypothesis, and obtain initial estimates of the channel parameters; The channel estimation module is used to update the channel parameter estimates during the iteration process using the pseudo-pilot symbol sequence provided by the data detection and regeneration module through the expectation-maximization algorithm. The data detection and regeneration module is used to process the received signal using the channel parameters updated by the channel estimation module, and to generate a new round of pseudo-pilot symbol sequence through an adaptive hybrid decision regeneration mechanism. The iteration control module is used to control the inner iteration loop and determine whether the iteration meets the convergence condition or reaches the maximum number of iterations. The final decision module is used to compare the log-likelihood values ​​of each hypothesis after all candidate schemes have been processed, and output the final modulation and coding recognition result and the corresponding decoded information bits.

6. A computer terminal device, characterized in that, include: One or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors perform the steps of the method as described in any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-4.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-4.

Images