Polarization code adaptive decoding method based on path distinction degree judgment

Abstract

The invention belongs to the technical field of communication, and relates to a polarization code adaptive decoding method based on path discrimination judgment. According to the method, a judgment index used for representing a current decoding state is constructed based on path metric information in an initial stage of a decoding process, and a self-adaptive decoding mode is selected and locked according to the judgment index. The method comprises the steps that a decoding tree is traversed under a fast simplified serial Cancellation List (FSSCL) decoding framework, when mixed nodes are traversed, a corresponding self-adaptive list management strategy is executed according to a locked decoding mode, and the strategy comprises a dynamic pruning strategy based on the path distinguishing degree and a list adjusting strategy based on verification feedback. By adopting a matched list management mode in different decoding states, the method can effectively reduce the average decoding complexity while ensuring the decoding performance, and is suitable for polar code decoding scenes under different channel conditions.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication channel coding and decoding technology, specifically relating to a serial cancellation list (SCL) decoding method for polar codes, and more particularly to an adaptive decoding method based on path discrimination decision, which can be applied to various decoding frameworks such as fast and simplified serial cancellation list decoding. Background Technology

[0002] Polar codes, as a channel coding scheme that can theoretically reach the Shannon limit, have been applied to next-generation wireless communication systems. In practical decoding, the SCL decoding algorithm achieves a good trade-off between decoding performance and implementation complexity by introducing a finite number of candidate paths, and is therefore widely used.

[0003] To further reduce decoding complexity and improve decoding speed, researchers proposed the Fast Simplified Successive Cancellation List (FSSCL) decoding framework. This framework leverages the node structure characteristics in the polar code factor graph to classify some nodes in the decoding tree as special nodes that can be processed quickly, thereby significantly reducing computational cost without significantly sacrificing decoding performance. Under the FSSCL framework, the decoding process typically uses nodes as the basic processing unit and performs a depth-first traversal.

[0004] On the other hand, to adapt to different channel conditions and decoding difficulties, various adaptive SCL decoding methods have been proposed in previous studies. One type of method uses a check feedback mechanism to gradually adjust the list size to reduce the average decoding complexity under favorable conditions; another type of method analyzes the distribution characteristics of path metrics and dynamically prunes candidate paths to reduce the retention of redundant paths. These methods each have their advantages in different application scenarios, but their designs are mostly based on bit-level or whole-frame-level decoding processes, and do not fully consider the impact of the node structure in the FSSCL framework on the balance between decoding complexity and performance.

[0005] Therefore, under the FSSCL decoding framework, how to utilize the path information available during the decoding process to adaptively select a decoding strategy suitable for the current decoding state, and further reduce the overall decoding complexity while ensuring decoding performance, remains to be solved. Summary of the Invention

[0006] To address the aforementioned problems, this invention proposes an adaptive polar code decoding method based on path discrimination decision. This method analyzes the statistical characteristics of path metrics during the decoding process, constructs a decoding state decision index, and adaptively selects different decoding strategies for hybrid nodes within the FSSCL framework, thereby achieving a reasonable balance between decoding performance and complexity under different decoding states.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] An adaptive decoding method for polar codes based on path discrimination decision is characterized by the following steps:

[0009] Step S1: Decoding initialization and calculation of decoding state decision index

[0010] The received polar code signal is initialized to obtain the corresponding Log-Likelihood Ratio (LLR) sequence. During decoding, a decision index M is constructed to characterize the current decoding state based on the path metric values ​​of each path in the current path list.

[0011] The decision index M is composed of the statistical characteristics of the difference in path metrics, which is used to reflect the degree of distinction between candidate paths in the current decoding stage. Its value corresponds to the actual channel conditions, but explicit calculation of the channel signal-to-noise ratio is not required.

[0012] Step S2: Adaptive Decoding Mode Decision and Locking

[0013] The decision index M is compared with the preset mode switching threshold M. th Compare the results and select the decoding mode accordingly:

[0014] When M ＜ M th When choosing a path discrimination-driven adaptive decoding mode;

[0015] When M≥M th When selecting, choose the adaptive decoding mode driven by verification feedback.

[0016] The decoding mode remains unchanged during the decoding process of the current codeword, and under the fast simplified serial cancellation list decoding framework, it takes effect uniformly on all hybrid nodes that need to use the SCL mechanism for decoding, in order to guide the decoding process of subsequent hybrid nodes.

[0017] Step S3: Perform node traversal decoding within the FSSCL decoding framework.

[0018] Within the FSSCL decoding framework, a depth-first traversal is performed on the decoding tree, and a global path list is maintained. For different node types encountered during the traversal, the following processing is performed:

[0019] Step S31: Fast decoding of special nodes

[0020] When a special node that can be quickly decoded is encountered during traversal, the corresponding fast decoding unit is invoked to independently calculate the decoding result of the covered bits of the node for each path in the global path list, and the path metric and the sequence of decoded bits are updated, while the size of the path list remains unchanged.

[0021] Step S32: Adaptive Decoding of Hybrid Nodes

[0022] When a mixed node that cannot be decoded quickly is encountered, the adaptive decoding kernel is invoked, and the corresponding decoding strategy is executed according to the decoding mode locked at the start of the current codeword decoding in step S2.

[0023] In the path discrimination-driven adaptive decoding mode, a path expansion operation is performed on the information bits within the hybrid node, the path metric value of the expanded candidate path is calculated, and the candidate path is selectively retained or pruned based on the comparison result of the difference between the adjacent path metrics after sorting and the pruning threshold, thereby achieving adaptive adjustment of the list size without exceeding the maximum list size.

[0024] In the verification feedback-driven adaptive decoding mode, the adaptive list management strategy is locked at the start of the current codeword decoding. When traversing to a hybrid node, list decoding processing is performed on the information bits within the hybrid node according to this strategy. Initially, a small list size is used for decoding, and after decoding, the decoding result is judged using verification information. If the verification passes, the current decoding result is accepted; if the verification fails, the list size is increased while maintaining the decoding strategy, and the corresponding list decoding processing is re-executed until the verification passes or the preset maximum list size limit is reached.

[0025] Step S33: Path list update and transfer

[0026] After decoding hybrid or special nodes, the updated path list is passed to the FSSCL decoding framework for processing subsequent nodes.

[0027] Step S4: Output the decoding result

[0028] After traversing all nodes of the decoding tree, the path with the best path metric is selected from the final path list and output as the decoding result.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. Overall optimization for the FSSCL framework

[0031] This invention introduces an adaptive decoding mechanism only in hybrid nodes, giving full play to the fast decoding advantage of special nodes in the FSSCL framework.

[0032] 2. Decoding State-Driven Strategy Selection Mechanism

[0033] By constructing a decoding state decision index based on path metric difference, an effective characterization of the degree of differentiation of candidate paths in the current decoding stage can be achieved.

[0034] 3. Adaptive balance between complexity and performance

[0035] The system automatically selects the appropriate adaptive decoding mode based on different decoding states, effectively reducing the average decoding complexity while ensuring decoding performance.

[0036] 4. User-friendly engineering implementation

[0037] The decoding mode is locked at the start of a decoding process and remains unchanged throughout the codeword decoding process, avoiding the control overhead caused by frequent strategy switching and making it easy to implement and verify in existing software simulation platforms. Attached Figure Description

[0038] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration:

[0039] Figure 1 This is a schematic diagram of the overall process of the method of the present invention;

[0040] Figure 2 This is a schematic diagram illustrating the relationship between node processing and adaptive decoding kernel calls in the FSSCL decoding framework.

[0041] Figure 3 A flowchart illustrating the adaptive decoding mode driven by path discrimination.

[0042] Figure 4 This is a flowchart illustrating the adaptive decoding mode driven by verification feedback. Detailed Implementation

[0043] 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. This embodiment uses a polar code length N = 1024, a code rate R = 1 / 2, and a maximum list size L. max This invention is illustrated using an example, but is not limited to these specific parameters.

[0044] See Figure 1The method of this invention operates within the FSSCL framework. When the decoding process enters a node that requires path expansion using the SCL mechanism, the following adaptive decoding steps are initiated:

[0045] Step 1: Calculation of the decoding state decision index

[0046] In the current decoding stage, let the size of the active path list inherited from the previous decoding stage be L. Path expansion is performed on the current information bits, resulting in 2L candidate paths. For the i-th candidate path, its path metric PM... i Update as follows:

[0047]

[0048] Among them, PM parent This is the path metric for this node at its parent node within the current path. For the hard decision result of this bit under the corresponding path, LLR i Let be the log-likelihood ratio of that bit.

[0049] Sort all candidate paths in ascending order of path metric values ​​to obtain an ordered sequence of path metrics:

[0050] PM (1) ≤PM (2) ≤…≤PM (2L)

[0051] Step 2: Construction of the Path Metric Difference Decision Index

[0052] Calculate the path metric difference between adjacent paths based on the sorted path metric sequence:

[0053] ΔPM i =PM (i+1) -PM (i) i = 1, 2, ..., 2L-1

[0054] Select the top K optimal paths after sorting (K is a preset constant, preferably 2 to 4), and construct the decoding state decision index:

[0055]

[0056] The decision index M is used to characterize the degree of distinction between the optimal candidate paths in the current decoding stage.

[0057] Step 3: Decoding State Decision and Policy Locking

[0058] Preset decoding state switching threshold M th The currently calculated decision index M is compared with the threshold M. th Comparison:

[0059] When M , M th When the path discrimination is weak in the current decoding state, the path discrimination-driven adaptive decoding mode is selected.

[0060] When M≥M th If the path is determined to be highly distinguishable under the current decoding state, the adaptive decoding mode driven by verification feedback is selected.

[0061] The selected decoding mode is determined at the start of decoding the current codeword and remains unchanged throughout the decoding process. Under the FSSCL framework, it applies uniformly to all subsequent hybrid nodes that need to adopt the SCL mechanism.

[0062] Step 4: Decoding Adaptive List Management under the FSSCL Framework

[0063] See Figure 2 The FSSCL decoder traverses the decoder tree in a depth-first manner. During the traversal, it performs the following processing based on the node type.

[0064] (I) Fast Decoding of Special Nodes

[0065] When encountering a special node that can be quickly decoded (including fully frozen nodes, fully information nodes, repeating code nodes, and single parity check nodes), the corresponding fast decoding unit is invoked. This unit directly calculates the decoding result of the node's covered bits without performing path expansion, and updates the path metric and the sequence of decoded bits, while keeping the size of the path list unchanged.

[0066] (II) Adaptive Decoding of Hybrid Nodes

[0067] When a mixed node that cannot be decoded quickly is encountered, the adaptive decoding kernel is invoked, and the corresponding strategy is executed based on the decoding mode that was locked at the start of the current codeword decoding.

[0068] 1) Path-discriminative driven adaptive decoding mode

[0069] See Figure 3 In this mode, for the 2L candidate paths and their path metrics PM obtained by expanding the current hybrid nodes, an adaptive list management process based on the path metric difference is performed.

[0070] First, based on the path metric difference vector ΔPM i Based on the statistical characteristics, adaptively construct the pruning threshold, including but not limited to construction methods based on the mean, standard deviation, or combinations thereof, and calculate the upper limit threshold T used in the path selection process. high and lower limit threshold T low First, calculate ΔPM. iThe mean and standard deviation of a vector, denoted as ΔPM i The mean of the vector is PM mean The standard deviation is PM std ,but

[0071] T high =PM mean

[0072] T low =PM mean -3*PM std

[0073] The above-described threshold construction method is a preferred embodiment. Those skilled in the art can construct thresholds using other statistical quantities or functional forms without departing from the spirit of the present invention.

[0074] Then, the paths are selected in descending order of path metrics:

[0075] For the difference between the first L paths after sorting, if ΔPM i ≤T high If ΔPM, then both adjacent paths are retained; i＞ T high If the current path is not selected, then only the current path will be retained, and the path selection process will be terminated.

[0076] For the difference between the L paths after sorting, if ΔPM i ≤T low If ΔPM, then both adjacent paths are retained; i >T low If the current path is not selected, then only the current path will be retained, and the path selection process will be terminated.

[0077] During the path selection process, if the number of paths to retain reaches the maximum list size L... max If the selection fails, the selection process stops. The final output is the updated list of paths as the decoding result for the current hybrid node.

[0078] 2) Verification feedback-driven adaptive decoding mode

[0079] See Figure 4 In this mode, the decoding process treats the entire data frame as a processing unit and employs an adaptive list management strategy based on check feedback. Decoding begins with an initial list size L=1, using the current L as a global list size constraint, and performs complete SCL decoding on the hybrid node. After decoding, check information (e.g., CRC) attached to the data frame is used to verify each path in the final surviving path list.

[0080] If one or more paths pass the CRC checksum, output the path with the highest reliability that passes the CRC checksum, accept the decoding result, and end the processing of the current node; if no path passes the CRC checksum, increase the list size according to the preset growth rule and re-decode until a path passes the CRC checksum or the maximum list size L is reached. max If the list size has reached L... max However, since no path has yet passed the CRC check, the path with the highest reliability is output and the current decoding ends.

[0081] Step 5: Return the FSSCL frame and output the decoding result.

[0082] Regardless of the decoding mode used to process the mixed nodes, an updated path list is output and returned to the FSSCL framework to continue decoding subsequent nodes. After traversing the entire decoding tree, the path with the optimal path metric is selected from the final path list as the final decoding result.

Claims

1. A polar code adaptive decoding method based on path discrimination decision, characterized in that, The specific steps are as follows: Step 1: Initialize the received polar code signal to obtain the corresponding log-likelihood ratio information, and construct a decision index M to characterize the current decoding state based on the path metric information and its statistical characteristics available in the initial stage of decoding. Step 2: Compare the decision index M with the preset strategy switching threshold M th A comparison is made, and an adaptive decoding mode is selected and locked at the start of decoding the current codeword, the adaptive decoding mode remaining unchanged throughout the decoding process; Step 3: Perform node-level traversal of the decoding tree under the fast simplified serial cancellation list decoding framework. When a special node that can be quickly decoded is traversed, fast decoding processing is performed. When a mixed node is traversed, the corresponding adaptive list management decoding strategy is executed on the mixed node according to the locked adaptive decoding mode. Step 4: After decoding all nodes, select the path with the best path metric from the final path list as the decoding result output.

2. The polar code adaptive decoding method based on path discrimination decision according to claim 1, characterized in that: In step 1, the decision index is used to characterize the degree of distinction between candidate paths and corresponds to the actual channel conditions without explicitly estimating the channel signal-to-noise ratio.

3. The polar code adaptive decoding method based on path discrimination decision according to claim 1, characterized in that: In the method of claim 1, when the decision index is less than the policy switching threshold, the decoding process further includes an adaptive decoding mode driven by path discrimination; in this mode, by analyzing the path metric ranking results of candidate paths after path expansion, and based on the statistical characteristics of the difference in metric between adjacent paths, candidate paths are adaptively retained or pruned to achieve dynamic adjustment of the list size.

4. The polar code adaptive decoding method based on path discrimination decision according to claim 1, characterized in that: In the method of claim 1, when the decision index is greater than or equal to the strategy switching threshold, the decoding process further includes executing a verification feedback-driven adaptive decoding mode; in this mode, while keeping the decoding strategy unchanged, the decoding process is repeatedly executed by gradually adjusting the list size and combining the verification information of the decoding result until the verification passes or the maximum list size limit is reached.

5. The polar code adaptive decoding method based on path discrimination decision according to claim 1, characterized in that: The adaptive decoding mode is determined at the start of decoding the current codeword and remains unchanged throughout the decoding process to avoid the control overhead caused by frequently switching decoding strategies.

Images