A soft decision decoding method and device based on symbol-level candidate set and block-level constraint, a chip and a storage medium

Abstract

The application discloses a soft decision decoding method and device based on symbol-level candidate set and block-level constraint, a chip and a storage medium, which forms a symbol-level candidate set by retaining multiple candidate symbols and corresponding metric values in each symbol period; meanwhile, a plurality of received symbols are divided into symbol blocks containing data symbols and redundant symbols, candidate combinations are generated based on each symbol candidate set and comprehensive metrics are calculated, combinations satisfying the constraint are screened by using a preset constraint relationship between the data symbols and the redundant symbols, and finally target candidate combinations are selected according to the comprehensive metrics to recover the sending data; meanwhile, the symbol-level candidate set and the block-level constraint replace high-complexity bit-level soft decoding, on the premise of keeping low calculation, low storage occupation and low power consumption, efficient and reliable decoding suitable for Internet of Things chips is realized, and the problem that the traditional soft decoding has high complexity in the prior art under a complex channel is effectively solved.

Description

Technical Field

[0001] This invention relates to the fields of wireless communication and communication chip technology, specifically to a soft-decision decoding method, apparatus, chip, and storage medium based on symbol-level candidate sets and block-level constraints. Background Technology

[0002] In low-power IoT communication systems, common physical layer modulation methods include FSK / GFSK, MFSK, chirp spread spectrum modulation, LoRa-type spread spectrum modulation, DSSS, PPM / UWB, PSK / QAM, and OFDM subcarrier modulation. Existing receivers typically demodulate within each symbol period, outputting the most probable symbol value as the hard decision result. To improve reception performance under low signal-to-noise ratio conditions, the system employs channel coding methods such as convolutional codes, Turbo, LDPC, or Polar, combined with bit-level LLR or path metrics to achieve soft decoding. Furthermore, existing multi-level modulation receiver schemes generally separate symbol demodulation from channel decoding; after the front-end outputs a single hard decision symbol, the subsequent stage performs bit error correction.

[0003] However, traditional bit-level soft decoding has high implementation complexity, storage overhead, iteration count, and power consumption, making it unsuitable for IoT chips that require low cost, low power consumption, and long battery life. Furthermore, in multipath, impulse interference, frequency offset, Doppler, or dynamic channel environments, the received peak value of a single symbol may not correspond to the actual transmitted symbol, which can lead to the hard decision error being solidified. Consequently, the subsequent error correction capability is also limited, making it impossible to reliably recover the correct symbol masked by interference under low complexity conditions. Summary of the Invention

[0004] The present invention aims to provide a soft decision decoding method, device, chip and storage medium based on symbol-level candidate set and block-level constraints, which effectively solves the problem of high complexity of traditional soft decoding under complex channels.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a soft-decision decoding method based on symbol-level candidate sets and block-level constraints, comprising the following steps:

[0006] S1: Perform front-end processing on the received signal to obtain the processed received signal;

[0007] S2: Demodulate the processed received signal in each symbol period to generate multiple candidate symbols and corresponding metric values, forming a candidate set for each received symbol;

[0008] S3: Divide several received symbols into symbol blocks according to preset rules. Each symbol block includes N data symbols and R redundant symbols.

[0009] S4: Generate candidate combinations based on the candidate sets of each symbol in the symbol block, and calculate the comprehensive metric for each candidate combination;

[0010] S5: Determine whether the candidate combination satisfies the preset constraint relationship between the data symbol and the redundant symbol, and select the target candidate combination from the candidate combinations that satisfy the preset constraint relationship to restore the transmission data according to the comprehensive metric.

[0011] Preferably, in S2, the candidate symbols are sorted according to the correlation peak, FFT peak, matched filter output, Euclidean distance or likelihood value, and the top K are retained as the candidate set; K is a fixed value or is adaptively adjusted according to the signal-to-noise ratio, peak difference, frequency offset indication or multipath indication.

[0012] Preferably, in S3, the preset rules include continuous segmentation, interleaved segmentation, or cross-resource segmentation;

[0013] The symbols within the symbol block are interleaved and mapped to discontinuous time positions, frequency positions, chirp symbol positions, OFDM resource units, or pulse positions.

[0014] Preferably, in S5, the preset constraint relationship includes one or more of the following: parity check relationship, modulo M summation relationship, finite field linear check relationship, matrix check relationship, lookup table constraint relationship, local CRC or short check code relationship, differential constraint relationship, phase constraint relationship or frequency constraint relationship.

[0015] Preferably, in S4, when generating the candidate combination, one or more symbol index offsets q are applied to the candidate symbols to compensate for the symbol index offset caused by frequency offset, Doppler frequency shift or sampling frequency offset;

[0016] The symbol index offset q is a value in the set of integers.

[0017] Preferably, multiple consecutive symbol blocks share the same symbol index offset q, or the symbol index offsets q of adjacent symbol blocks satisfy a continuity constraint, or different symbols within the same symbol block use segmented consecutive symbol index offsets q.

[0018] Preferably, in S5, if the candidate combination with the maximum metric does not satisfy the preset constraint relationship, then the combination that satisfies the preset constraint relationship and has the best comprehensive metric is selected from the non-maximum metric candidates as the target candidate combination; if there is no candidate combination that satisfies the preset constraint relationship, then the K value is increased, the symbol index offset q is changed, or a low bit rate enhancement mode is used.

[0019] On the other hand, the present invention proposes a soft-decision decoding device based on symbol-level candidate sets and block-level constraints, comprising:

[0020] The front-end processing module is used to perform front-end processing on the received signal to obtain the processed received signal.

[0021] The candidate set generation module is used to demodulate the processed received signal in each symbol period, generate multiple candidate symbols and corresponding metric values, and form a candidate set for each received symbol.

[0022] The symbol block partitioning module is used to divide a number of received symbols into symbol blocks according to preset rules. The symbol blocks include N data symbols and R redundant symbols.

[0023] The candidate combination generation module is used to generate candidate combinations based on the candidate sets of each symbol in the symbol block, and to calculate the comprehensive metric for each candidate combination.

[0024] The constraint decision module is used to determine whether the candidate combination satisfies the preset constraint relationship between the data symbol and the redundant symbol, and selects the target candidate combination to restore the transmission data based on the comprehensive metric among the candidate combinations that satisfy the preset constraint relationship.

[0025] On the other hand, the present invention proposes a chip including a processor and a memory, wherein the processor is configured to execute a computer program stored in the memory to implement the soft decision decoding method based on symbol-level candidate sets and block-level constraints as described above.

[0026] On the other hand, the present invention proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the soft-decision decoding method based on symbol-level candidate sets and block-level constraints as described above.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] This invention forms a symbol-level candidate set by retaining multiple candidate symbols and corresponding metric values ​​in each symbol period. Simultaneously, it divides several received symbols into symbol blocks containing data symbols and redundant symbols. Based on each symbol candidate set, candidate combinations are generated and a comprehensive metric is calculated. Then, a preset constraint relationship between data symbols and redundant symbols is used to filter combinations that meet the constraints. Finally, the target candidate combination is selected based on the comprehensive metric to recover the transmitted data. This avoids the problem of premature loss of correct symbol information in hard decision-making scenarios, significantly improving decoding sensitivity and robustness in low signal-to-noise ratio, multipath, and frequency offset / Doppler scenarios. Furthermore, by replacing the highly complex bit-level soft decoding with a symbol-level candidate set and block-level constraints, it achieves efficient and reliable decoding suitable for IoT chips while maintaining low computational load, low storage footprint, and low power consumption, effectively solving the problem of high complexity in traditional soft decoding under complex channels. Attached Figure Description

[0029] Figure 1 This is the overall flowchart of the receiver of the present invention;

[0030] Figure 2 This is a graph showing the generation of the TopK candidate set for a single symbol in this invention.

[0031] Figure 3 This is the N+R symbol block structure and interleaving diagram of the present invention;

[0032] Figure 4 This is the block-level candidate combination search graph of the present invention;

[0033] Figure 5 This is a diagram illustrating the TopK recovery mechanism under multipath conditions according to the present invention.

[0034] Figure 6 This is a graph showing the symbol index offset q search and trajectory constraint of the present invention;

[0035] Figure 7 This is a unified symbol-level candidate representation diagram under various modulation methods of the present invention;

[0036] Figure 8 This is a diagram illustrating the low-complexity chip implementation architecture of the present invention.

[0037] Figure 9 The flowchart shows the soft-decision decoding method based on symbol-level candidate sets and block-level constraints of the present invention.

[0038] Figure 10 This is a block diagram of the soft decision decoding system based on symbol-level candidate sets and block-level constraints of the present invention. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0040] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a specific posture. If the specific posture changes, the directional indicators will also change accordingly.

[0041] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the use of "and / or" or "and / or" throughout the text includes three parallel solutions. For example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0042] This embodiment proposes a chip and a computer-readable storage medium storing a computer program, including a processor and a memory. The processor is configured to execute the computer program stored in the memory to implement the soft-decision decoding method based on symbol-level candidate sets and block-level constraints as described below.

[0043] Please see Figures 1-9 As shown, this invention provides a soft-decision decoding method based on symbol-level candidate sets and block-level constraints. The receiving end generates a candidate symbol set and its metric value for each received symbol, and then divides the multiple symbols into symbol blocks. Each symbol block includes multiple data symbols and at least one redundant or discriminant symbol. The receiving end forms candidate combinations from the candidate sets of each symbol and filters them according to preset constraint relationships between data symbols and redundant symbols. Among the candidate combinations that satisfy the constraint relationships, the combination with the minimum total cost or the maximum total likelihood is selected as the target combination, thereby recovering the transmitted data. Specifically, the method includes the following steps:

[0044] S1: Perform front-end processing (synchronization, coarse frequency offset correction, symbol timing) on ​​the received signal to obtain the processed received signal; wherein, synchronization, coarse frequency offset correction, symbol timing, and necessary front-end processing can be implemented by one or more combinations according to the modulation method and chip architecture, specifically including but not limited to the following methods:

[0045] Synchronization utilizes preambles, pilots, training sequences, energy detection, correlation detection, chirp up / down detection, or codeword correlation peaks to detect the start position, which is used to obtain the packet start position, symbol boundary, or coarse timing.

[0046] Coarse frequency offset correction refers to estimating the carrier frequency offset by preamble FFT bin offset, or by phase difference between adjacent symbols, or by phase rotation of pilot / training sequences, and then performing digital compensation through numerically controlled oscillator (NCO). Its purpose is to first correct large frequency offsets to a demodulation range and reduce the probability of candidate set misalignment.

[0047] Symbol timing is determined by the position of the maximum correlation peak, sliding window correlation, leader period consistency, or peak position of the training sequence to ensure that each symbol period corresponds to the correct reception window.

[0048] Necessary front-end processing includes automatic gain control (AGC), DC offset removal, in-phase / quadrature IQ signal imbalance correction, digital downconversion, low-pass filtering, decimation, and normalization, which are used to improve the stability of subsequent TopK candidate generation.

[0049] S2: Demodulate the processed received signal in each symbol period to generate multiple candidate symbols and corresponding metric values, forming a candidate set C_i={c_i,1,c_i,2,...,c_i,K} for each received symbol;

[0050] Demodulation is not limited to a specific modulation method; it is used to output multiple candidate symbols for each received symbol, rather than just a single maximum likelihood hard-decision symbol. Candidate symbols are sorted according to correlation peak, FFT peak, filter bank output, Euclidean distance, likelihood value, or other confidence metrics, as shown in the table below:

[0051] Table 1. Specific implementation examples of single-symbol demodulation and TopK candidate generation:

[0052] CSS / LoRa / Chirp Spread Spectrum FFT bin, chirp index, prime point index, multi-slope index After dechirp for each symbol, perform FFT or equivalent frequency search, and retain the TopK bins by peak amplitude / energy. MFSK / Multi-base FSK Frequency Index The energy of each frequency point is obtained using a filter bank, the Goertzel algorithm, or a multi-frequency correlator, and the Top K frequency points with the highest energy are retained. FSK / GFSK / GMSK / CPFSK Single-symbol candidates or short-sequence candidates Single-symbol or short-sequence candidates are formed from continuous phase trajectories, frequency trajectories, or matched filter outputs, and then sorted by path metric. QAM / PSK / APSK Constellation Points Based on the Euclidean distance between the receiver and the constellation points, the distance after phase / amplitude correction, or the likelihood value, retain the TopK constellation points. OFDM Subcarrier constellation point or resource unit symbol After equalization, TopK constellation candidates are generated at the receiver point for each subcarrier, and joint constraint search can be performed within a resource block. DSSS / Spreading Code Index Spread codewords and chip sequence candidates Correlate multiple spreading codes or codewords and retain the Top K codeword indices with the largest correlation peaks. UWB / PPM Pulse Time Position Index For possible pulse time locations, perform correlation or energy detection, and retain the Top K time locations with the largest energy / correlation peaks.

[0053] Candidate symbols are sorted according to correlation peak value, FFT peak value, matched filter output, Euclidean distance, or likelihood value, and the top K symbols are retained as the candidate set. K is a fixed value, or it can be adaptively adjusted according to signal-to-noise ratio, peak difference, frequency offset indication, and multipath indication. K is less than or equal to the total number of symbol states contained in the modulation symbol set. When K is equal to the total number of symbol states, the candidate set includes all symbol states and the corresponding metric values ​​for each symbol state.

[0054] In chirp spread spectrum or LoRa-type modulation, the receiver can perform dechirp processing on each symbol, followed by FFT or equivalent frequency search, retaining the Top K bins with the largest amplitudes as candidate symbols. In MFSK, the Top K frequency points with the largest energy are retained; in QAM / PSK, the Top K constellation points closest to the receiver or with the highest likelihood are retained; in PPM / UWB, the Top K time positions with the largest correlation energy are retained; in DSSS or code-indexed modulation, the Top K codeword indices with the largest correlation peaks are retained.

[0055] K can be a fixed value, such as 2, 4, 8, or 16, or it can be adaptively adjusted based on SNR, the difference between the largest and second-largest peaks, channel state, or preamble estimation results. When SNR is high and the confidence of the largest peak is high, only hard decision or Top2 can be used; under conditions of multipath, frequency offset, Doppler, or low SNR, Top4, Top8, or larger candidate sets can be enabled.

[0056] When the total number of modulation symbol states is less than or equal to K, all symbol states and their corresponding metrics can be saved. For example, when the spreading factor SF is equal to 3, one symbol corresponds to 8 possible states. When K is 8, the candidate set includes all 8 symbol states. The receiver performs a complete symbol-level soft decision based on the metrics of each state and the block-level constraints, rather than relying solely on the candidate set truncation.

[0057] For a symbol block consisting of 8 data symbols and 1 redundant symbol, the combination consisting of 9 Top 1 candidates can be verified first. If the verification fails, several least reliable symbols are selected according to the symbol reliability, and their Top 2, Top 4 or Top 8 candidates are tried in turn. Invalid combinations are screened out using the redundant symbol constraint.

[0058] For a symbol block containing L = N + R symbols, if each symbol retains K candidates, the theoretical number of candidate combinations is K^L. This application reduces complexity through constraint-based fast filtering, metric sorting, path pruning, and early stopping.

[0059] One implementation is as follows: First, perform a hard decision based on the Top 1 candidate for each symbol to verify the symbol block constraints; if the verification passes, there is no need to enter a complex search; if it fails, combine the candidate metrics from high to low, prioritizing combinations with better total metrics, and use redundancy constraints for judgment. If a combination satisfies the constraints and its metric advantage exceeds the threshold, then output the combination; if multiple combinations satisfy the constraints, select the combination with the largest total correlation peak energy, the smallest total distance, or the largest total likelihood.

[0060] The comprehensive metric can be defined as a weighted sum of candidate symbol metrics, log-likelihood sum, energy difference, Euclidean distance sum, normalized reliability, etc. For FFT bins or MFSK frequencies, the metric can be peak amplitude or peak energy; for QAM constellations, the metric can be the Euclidean distance between the receiver and candidate constellation points; for correlated codewords, the metric can be the correlated peak value.

[0061] During chip implementation, only the TopK candidate indices and metric values ​​for each symbol can be cached, without needing to store the complete FFT or the complete likelihood array. Candidate search can be implemented using finite state machines, lookup tables, pipelines, or small-scale combinational logic.

[0062] In multipath channels, the superposition of the main path and the reflection path may cause a decrease in the peak value of the true symbol, or generate secondary peaks in nearby bins, nearby frequency points, or nearby time positions. Hard decision only selects the maximum peak, which can easily lead to the misinterpretation of secondary peaks as true symbols.

[0063] This embodiment retains the TopK candidates, ensuring that even if the true symbol drops from Top1 to Top4 or Top8, it can still be preserved in the candidate combination search. Then, by filtering through symbol block constraints to select combinations that satisfy redundancy relationships and have reasonable metrics, the correct symbol can be recovered.

[0064] Taking chirp spread spectrum as an example, static or dynamic multipath may cause perturbations in the FFT bin ranking. If the true bin is still within the TopK, this application can select it from the candidate list through block-level verification. Simulation verification shows that, compared with hard decision, Top4 / Top8 symbol-level soft decision can improve sensitivity in AWGN environment and significantly alleviate the problem of increased packet error rate caused by multipath in multipath environment.

[0065] S3: Divide several received symbols into symbol blocks according to preset rules. Each symbol block includes N data symbols and R redundant symbols. The preset rules include continuous block division, interleaved block division, or cross-resource block division. The symbols in the symbol block are interleaved and mapped to non-continuous time positions, frequency positions, chirp symbol positions, OFDM resource units, or pulse positions.

[0066] Multiple consecutive symbol blocks share the same symbol index offset q, or the symbol index offsets q of adjacent symbol blocks satisfy a continuity constraint, or different symbols within the same symbol block use segmented consecutive symbol index offsets q.

[0067] Examples of preset rules are as follows:

[0068] Eight data symbols plus one redundant symbol form a symbol block with a code rate of 8 / 9;

[0069] Seven data symbols plus one discrimination / redundancy symbol form a symbol block with a code rate of 7 / 8;

[0070] Six data symbols plus one discrimination / redundancy symbol form a symbol block with a code rate of 6 / 7;

[0071] Five data symbols plus one discrimination / redundancy symbol form a symbol block with a code rate of 5 / 6;

[0072] Four data symbols plus one discrimination / redundancy symbol form a symbol block with a code rate of 4 / 5;

[0073] Four data symbols + two redundant symbols, or N data symbols + R redundant symbols, are suitable for enhanced modes with stronger error correction or lower speeds.

[0074] Symbols in the same symbol block can be transmitted directly and continuously, or they can be distributed to different time positions, frequency positions, chirp symbol positions, OFDM resource units, codeword positions, or pulse positions through interleaving.

[0075] In this embodiment, the symbol block consists of multiple data symbols and at least one redundant symbol. For example, a typical implementation is a symbol block with 4 data symbols and 1 discriminant symbol, i.e., a code rate of 4 / 5. Alternatively, it can use 5 data symbols and 1 discriminant symbol, 6 data symbols and 1 discriminant symbol, 7 data symbols and 1 discriminant symbol, or N data symbols and R redundant symbols.

[0076] Redundant symbols and data symbols can satisfy any one or more of the following constraints:

[0077] Parity check relationships, such as XORing some or all bits of multiple symbols to obtain the parity symbol;

[0078] Modulo M summation relationship, for example, redundant symbols are equal to the sum of multiple data symbol indices modulo M;

[0079] Lookup table constraints, which are constraints that use multiple data symbols to look up a table to obtain redundant symbols;

[0080] Local CRC or short check code relationship;

[0081] Constraints related to modulation symbol index, symbol difference, phase difference, or frequency difference.

[0082] The above constraints are not limited to a single redundant symbol, nor are they limited to binary checksums. For multi-level modulation where a symbol carries multiple bits or multiple states, redundant symbols can be generated directly at the symbol level to reduce the complexity of bit-level complex software decoding.

[0083] To avoid consecutive channel impairments concentrating on the same symbol block, this application interleaves data symbols and redundant symbols at the transmitting end, distributing symbols within the same symbol block across different time positions, frequency positions, chirp symbol positions, OFDM resource elements, or other modulation resources. After the receiving end completes candidate set generation, it performs deinterleaving according to the same interleaving rules before proceeding to symbol block-level candidate search.

[0084] Interleaving methods can include fixed interleaving, pseudo-random interleaving, frame-length adaptive interleaving, channel-state-selective interleaving, or interleaving across multiple blocks. Interleaving depth can be configured based on rate, delay, mobility, and target anti-jamming capability.

[0085] Table 2 Applicable Modulation Methods:

[0086] CSS / LoRa / Chirp Spread Spectrum FFT bin, chirp index, prime point index, multi-slope index FFT peak, correlation peak TopK bin+N+R block constraints; supports q offset anti-Doppler. MFSK / Multi-base FSK Frequency Index Frequency energy, filter bank output, correlation peak TopK frequency points + symbol-level verification; can compensate for overall frequency offset. FSK / GFSK / GMSK / CPFSK Single-symbol candidates or short-sequence candidates Phase trajectory, frequency trajectory, matched filter metric Short sequence candidates are replaced with single-symbol candidates, combined with block-level constraints. QAM / PSK / APSK Constellation Points Euclidean distance, likelihood value, and distance after phase / amplitude correction TopK constellation points + block-level verification; phase rotation / remapping assumptions can be introduced. OFDM Subcarrier constellation point or resource unit symbol Soft metrics and Euclidean distances for each subcarrier Candidate combinations and block-level constraints are performed within multiple subcarriers or resource blocks. DSSS / Spread Code Index Modulation Spread codeword index, chip sequence candidate Correlation peaks and codeword matching degree TopK codeword candidates + local symbol block constraints. UWB / PPM Pulse Time Position Index Time-related peaks and energy detection results TopK time position + symbol block constraint; resists multipath / pulse position disturbances. Code Index / Sequence Index / Multi-user Codeword Codeword, sequence, or user index Correlation peaks, matching degree, probability measure The decision is made jointly using a candidate codeword list and N+R constraints.

[0087] Table 3 Modulation Examples:

[0088] Example 1: CSS / LoRa class modulation software decoding After receiving symbols undergo dechirp and FFT, the Top 4 / Top 8 bins and peak metrics are retained; every 4 data symbols and 1 redundant symbol form a symbol block; candidate combinations are selected for the correct bin through symbol-level verification. Example 2: Multipath Resistant Decoding Multipath propagation causes erroneous secondary peaks to become Top1, while the true bins drop to Top4 / Top8; hard decision fails. This method utilizes TopK candidates and block-level constraints to recover the true bins, thereby reducing the sensitivity loss caused by multipath propagation. Example 3: Fixed Doppler / Fixed Residual Frequency Offset When a fixed frequency offset causes multiple symbols to shift by one or more bins, the entire packet or multiple symbol blocks share q; for example, when SF12 and BW=125kHz, 30Hz corresponds to approximately 0.98bin, and selecting q=-1 can correct it back to the true bin. Example 4: ramp Doppler / frequency drift When the frequency offset gradually changes in the payload, the q of adjacent symbol blocks satisfies the continuity constraint, for example, q=0→-1→-2; if necessary, split / burst q is applied to the block boundary. Example 5: QAM / MFSK and other modulation generalization MFSK retains the TopK frequency point, QAM retains the TopK constellation point, DSSS retains the TopK codeword, and UWB / PPM retains the TopK time position; all can use N+R symbol block constraints and interleaving for joint decision-making.

[0089] S4: Generate candidate combinations based on the candidate sets of each symbol in the symbol block, and calculate the comprehensive metric for each candidate combination; when generating the candidate combination, apply one or more symbol index offsets q to the candidate symbols to compensate for symbol index offsets caused by frequency offset, Doppler frequency shift or sampling frequency offset; the symbol index offset q is a value in the set of integers.

[0090] When residual frequency offset, Doppler shift, sampling frequency offset, or clock drift are present, the candidate positions of multiple symbols may shift overall. For example, in chirp spread spectrum, the FFT peak may shift overall by one or more bins; in MFSK, the frequency candidates may shift overall; in PPM, the time position may shift systematically; and in QAM / PSK, the candidate constellation may undergo phase rotation or overall transformation.

[0091] To address this issue, this application introduces a symbol index offset hypothesis q in the candidate combination search. For indexed symbols, the receiver can transform the observed candidate symbol observed Candidate into corrected Candidate = observed Candidate + q, where q can be an integer in the range of -Q to +Q, for example, q∈{-3,-2,-1,0,+1,+2,+3}.

[0092] For fixed frequency offset or fixed Doppler, multiple consecutive symbols or multiple symbol blocks can share the same q; for frequency offset variations in the form of ramp, the q of adjacent symbol blocks can gradually change according to the continuity constraint; for burst or boundary scenarios, different symbols within the same symbol block can use segmented continuous q, for example, the first few symbols use q0, and the last few symbols use q0±1.

[0093] The receiver can first perform fractional tracking to bring the peak value back to the bin center as much as possible. If symbol block check, CRC, or confidence level determination indicates decoding failure, then an integer q search is initiated. The most reasonable q value is selected based on symbol block constraints, packet CRC, candidate metrics, and q-trajectory continuity. This method can handle situations where the signal amplitude is sufficient, but the overall symbol index is offset due to frequency offset / Doppler.

[0094] For example, in a chirp spread spectrum system with SF12 and BW=125kHz, a 30Hz residual frequency offset corresponds to approximately 0.98 bin offsets. Traditional hard-decision, Top4, and Top8 methods may fail to decode completely if integer offsets are not considered. This application uses the candidate offset assumption of q=-1 to correct the observed bins back to the true bins, and then combines block-level constraints to recover the data.

[0095] S5: Determine whether the candidate combination satisfies the preset constraint relationship between the data symbol and the redundant symbol, and select the target candidate combination from the candidate combinations that satisfy the preset constraint relationship to restore the transmission data according to the comprehensive metric.

[0096] If the candidate combination with the highest comprehensive metric does not satisfy the preset constraint, then the combination that satisfies the preset constraint and has the best comprehensive metric is selected from the remaining candidate combinations as the target candidate combination; if no candidate combination satisfies the preset constraint, then the K value is increased, the symbol index offset q is changed, or the low bit rate enhancement mode is switched. Afterwards, the symbol combinations that pass the decision undergo subsequent processing such as deinterleaving, demapping, dewhitening, and CRC check, and finally, the data is output.

[0097] The preset constraint relationships include one or more of the following: parity check relationship, modulo-M summation relationship, finite field linear check relationship, matrix check relationship, lookup table constraint relationship, local CRC or short check code relationship, differential constraint relationship, phase constraint relationship, or frequency constraint relationship, as shown in the table below:

[0098] Table 4. Block-level constraint judgment method:

[0099] Parity check XOR all or part of the bits of N data symbols to determine whether they are equal to a redundant symbol or a discrimination symbol. M plus Determine whether the redundant symbol r satisfies r = (s1 + s2 + ... + sN) mod M, where M can be the size of the symbol set or a preset modulus. Finite field linearity check Determine whether the data symbols and redundant symbols satisfy the linear check relationship on GF(2^m). Matrix verification Represent the candidate combination as a vector s, and determine whether H·s=0 or P·d=r holds true. Lookup table constraints Redundant symbols are obtained by looking up multiple data symbols in a table, and it is determined whether the candidate combination matches the preset table entry. Local CRC or short check Calculate the CRC or short check code for the local data recovered from the candidate combinations and compare it with the redundant symbols or check fields. Differential / phase / frequency constraints Determine whether the difference between adjacent symbols, phase difference, frequency difference, or symbol index difference meets the preset relationship.

[0100] On the other hand, this invention proposes a soft-decision decoding device based on symbol-level candidate sets and block-level constraints, such as... Figure 10 As shown, it includes:

[0101] The front-end processing module is used to perform front-end processing on the received signal to obtain the processed received signal.

[0102] The candidate set generation module is used to demodulate the processed received signal in each symbol period, generate multiple candidate symbols and corresponding metric values, and form a candidate set for each received symbol.

[0103] The symbol block partitioning module is used to divide a number of received symbols into symbol blocks according to preset rules. The symbol blocks include N data symbols and R redundant symbols.

[0104] The candidate combination generation module is used to generate candidate combinations based on the candidate sets of each symbol in the symbol block, and to calculate the comprehensive metric for each candidate combination.

[0105] The constraint decision module is used to determine whether the candidate combination satisfies the preset constraint relationship between the data symbol and the redundant symbol, and selects the target candidate combination to restore the transmission data based on the comprehensive metric among the candidate combinations that satisfy the preset constraint relationship.

[0106] This system is particularly suitable for implementation using a communication chip. A TopK candidate buffer module can be added to the demodulator output, and a symbol block candidate search module can be added before the FEC / deinterleaving module. The candidate buffer only stores candidate indices and a few bit metric values, without needing to store the complete spectrum, thus reducing storage overhead. To reduce power consumption, the following strategies can be adopted:

[0107] Hard decision and basic check first: TopK search is not started when the Top1 combination passes the block-level constraints and CRC.

[0108] Enable on demand: Only enable when SNR is low, multipath indicator, frequency offset indicator, Doppler indicator, or check failure is present;

[0109] Candidate number adaptive: K can be dynamically selected between 2, 4, 8, and 16;

[0110] q-gated search: q is searched only if there is an indication of an integer bin offset;

[0111] Early stopping: Once a candidate combination that satisfies the constraints and has sufficient metric advantage is found, the search for subsequent combinations is immediately stopped;

[0112] Hierarchical search: First try q=0 and the q with the highest probability, then expand to ±1, ±2, ±3.

[0113] Furthermore, each module in the above system is also used to implement other steps of the proposed soft-decision decoding method based on symbol-level candidate sets and block-level constraints, which will not be elaborated here.

[0114] Table 5. Chinese annotations for English abbreviations and terms:

[0115] TopK The top K candidates are retained after sorting by confidence, correlation peak, energy, likelihood, or distance. FFT Fast Fourier Transform (FFT) is used for frequency domain peak or bin search. CSS Chirp Spread Spectrum (CSP). LoRa A low-power wide-area network physical layer technology based on chirp spread spectrum. CFO Carrier Frequency Offset. Doppler Doppler shift is usually caused by relative motion or a moving reflection path. fractional tracking Fractional bin frequency offset tracking is used to correct small frequency offsets that do not cross integer bins. integerq The integer sign index offset q is used to correct the overall offset across bins or across indices. packet Data packets or physical layer frames. block A symbol block is a local constraint unit consisting of several data symbols and redundant / discriminative symbols. CRC Cyclic redundancy check is used to determine whether the entire packet or a part of the data is correct. OFDM Orthogonal frequency division multiplexing. DSSS Direct sequence spread spectrum. UWB Ultra-wideband. PPM Pulse position modulation.

[0116] The above description is merely an exemplary embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention specification and drawings under the technical concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A soft-decision decoding method based on symbol-level candidate sets and block-level constraints, characterized in that, Includes the following steps: S1: Perform front-end processing on the received signal to obtain the processed received signal; S2: Demodulate the processed received signal in each symbol period to generate multiple candidate symbols and corresponding metric values, forming a candidate set for each received symbol; S3: Divide several received symbols into symbol blocks according to preset rules. Each symbol block includes N data symbols and R redundant symbols. S4: Generate candidate combinations based on the candidate sets of each symbol in the symbol block, and calculate the comprehensive metric for each candidate combination; S5: Determine whether the candidate combination satisfies the preset constraint relationship between the data symbol and the redundant symbol, and select the target candidate combination from the candidate combinations that satisfy the preset constraint relationship to restore the transmission data according to the comprehensive metric; In S2, the candidate symbols are sorted according to the correlation peak, FFT peak, matched filter output, Euclidean distance or likelihood value, and the top K are retained as the candidate set; K is a fixed value or is adaptively adjusted according to the signal-to-noise ratio, peak difference, frequency offset indication or multipath indication; K is less than or equal to the total number of symbol states contained in the modulation symbol set; when K is equal to the total number of symbol states, the candidate set includes all symbol states and the metric value corresponding to each symbol state. For a symbol block consisting of 8 data symbols and 1 redundant symbol, first verify the combination consisting of 9 Top 1 candidates. If the initial Top 1 combination satisfies the constraints, it is directly output without initiating subsequent candidate expansion. If the verification fails, select several unreliable symbols according to the symbol reliability, and try their Top 2, Top 4 or Top 8 candidates in turn, and use the redundant symbol constraints to filter out illegal combinations. In S4, when generating the candidate combination, one or more symbol index offsets q are applied to the candidate symbols to compensate for the symbol index offset caused by frequency offset, Doppler frequency shift, or sampling frequency offset; the symbol index offset q is a value in the set of integers. Multiple consecutive symbol blocks share the same symbol index offset q, or the symbol index offsets q of adjacent symbol blocks satisfy a continuity constraint, or different symbols within the same symbol block use segmented consecutive symbol index offsets q; In S5, if the candidate combination with the maximum metric does not satisfy the preset constraint relationship, then the combination that satisfies the preset constraint relationship and has the best comprehensive metric is selected from the non-maximum metric candidates as the target candidate combination; if there is no candidate combination that satisfies the preset constraint relationship, then the K value is increased, the symbol index offset q is changed, or a low bit rate enhancement mode is used.

2. The soft-decision decoding method based on symbol-level candidate sets and block-level constraints according to claim 1, characterized in that, In S3, the preset rules include continuous segmentation, interleaved segmentation, or cross-resource segmentation; The symbols within the symbol block are interleaved and mapped to discontinuous time positions, frequency positions, chirp symbol positions, OFDM resource units, or pulse positions. The symbol block includes any of the following structures: Eight data symbols and one redundant symbol; Six data symbols and one redundant symbol; Four data symbols and one redundant symbol; Three data symbols and one redundant symbol; Four data symbols and two redundant symbols; Or N data symbols and R redundant symbols.

3. The soft-decision decoding method based on symbol-level candidate sets and block-level constraints according to claim 1, characterized in that, In S5, the preset constraint relationship includes one or more of the following: parity check relationship, modulo M summation relationship, finite field linear check relationship, matrix check relationship, lookup table constraint relationship, local CRC or short check code relationship, differential constraint relationship, phase constraint relationship or frequency constraint relationship.

4. A soft-decision decoding apparatus based on symbol-level candidate sets and block-level constraints for implementing the method as described in any one of claims 1-3, characterized in that, include: The front-end processing module is used to perform front-end processing on the received signal to obtain the processed received signal. The candidate set generation module is used to demodulate the processed received signal in each symbol period, generate multiple candidate symbols and corresponding metric values, and form a candidate set for each received symbol. The symbol block partitioning module is used to divide a number of received symbols into symbol blocks according to preset rules. The symbol blocks include N data symbols and R redundant symbols. The candidate combination generation module is used to generate candidate combinations based on the candidate sets of each symbol in the symbol block, and to calculate the comprehensive metric for each candidate combination. The constraint decision module is used to determine whether the candidate combination satisfies the preset constraint relationship between the data symbol and the redundant symbol, and selects the target candidate combination to restore the transmission data based on the comprehensive metric among the candidate combinations that satisfy the preset constraint relationship.

5. A chip, characterized in that, It includes a processor and a memory, the processor being configured to execute a computer program stored in the memory to implement the soft-decision decoding method based on symbol-level candidate sets and block-level constraints as described in any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the soft-decision decoding method based on symbol-level candidate sets and block-level constraints as described in any one of claims 1 to 3.

Images