An OTFS two-stage signal detection method based on global preliminary detection and local list refinement
Patent Information
- Application Number
- CN202610851843.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-12
- Publication Date
- 2026-09-01
AI Technical Summary
[0004]然而，仅依赖全局低复杂度初检测时，在部分低可靠且强耦合的位置处往往难以得到足够准确的软输出；若直接对整个时延–多普勒域网格实施高复杂度近似最优搜索，又会导致整体复杂度快速增长
[0027] (1) The present invention adopts a two-stage processing framework of global initial detection and local list fine detection, and performs selective fine detection only on low reliability positions, which can effectively control the average computational complexity.
Smart Images
Figure CN122679014A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of signal processing technology for wireless communication receivers, specifically relating to a two-stage signal detection method based on global initial detection and local list fine detection, which is applicable to system reliability detection scenarios of OTFS under dual-selective channel conditions. Background Technology
[0002] As next-generation wireless communication systems evolve towards higher frequency bands, greater mobility, and wider coverage, channels exhibit both significant time-varying characteristics and frequency selectivity. Traditional receiver methods designed for low Doppler conditions are prone to performance degradation in highly dual-selective scenarios. OTFS, by mapping information symbols to the delay-Doppler domain, can more naturally characterize dual-selective channels and has therefore become an important candidate waveform for high-mobility communication scenarios.
[0003] In OTFS systems, although the time-delay-Doppler domain equivalent channel typically exhibits structural local coupling characteristics, mutual interference still occurs between different symbol positions through the equivalent channel matrix. Therefore, reliable detection remains a critical issue in receiver design. Existing methods employ two main approaches: one uses linear equalization or approximate Gaussian inference for globally low-complexity detection, while the other employs message passing, iterative interference cancellation, or approximate Bayesian inference to improve detection performance.
[0004] However, relying solely on low-complexity initial detection globally often fails to yield sufficiently accurate soft outputs at locations with low reliability and strong coupling. Conversely, performing a high-complexity near-optimal search directly on the entire delay-Doppler domain grid leads to a rapid increase in overall complexity. Therefore, a detection method is needed that can maintain the advantage of low complexity in initial global detection while also enabling enhanced fine detection at a few critical locations. Summary of the Invention
[0005] The main objective of this invention is to overcome the shortcomings and deficiencies of the prior art and provide a two-stage signal detection method for OTFS based on global initial detection and local list fine detection. Through the two-stage processing framework, selective fine detection is performed only on locations with low reliability and significant local coupling, which can improve the detection reliability of key locations while controlling complexity.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] Firstly, this invention provides a two-stage OTFS signal detection method based on global initial detection and local list fine detection, including the following steps:
[0008] (1) Based on the OTFS received signal and the corresponding time-delay-Doppler domain equivalent channel matrix The first stage of global initial detection is performed to obtain initial detection information corresponding to each time delay-Doppler position. The initial detection information includes symbol-level initial metrics corresponding to each DD domain position. This also includes the initial log-likelihood ratio of each bit within each position. ;
[0009] (2) Calculate the reliability of each location based on the initial soft information output from the first stage obtained in step (1). And determine the set of locations to be re-inspected based on reliability. , where the set Includes components with reliability below a preset threshold The time delay-Doppler position;
[0010] (3) Based on the set of locations to be re-inspected obtained in step (2) The coupling relationships between different locations are calculated, and the coupling weights between locations are determined. And construct the threshold according to the coupling weight and graph. The comparison results are for the set of locations to be re-detected. Perform local segmentation to form one or more local detection blocks. ;
[0011] (4) For each local detection block obtained in step (3) Utilizing initial detection information from relevant locations outside the block Computation of soft symbol estimation And based on soft sign estimation Correcting local observations yields the corresponding local detection block. Local correction observations ;
[0012] (5) For the local detection block obtained in step (3) Based on the initial log-likelihood ratio obtained in step (1) Hard decision is performed on each bit within the local detection block to obtain a reference bit vector within the block; controlled bit flipping is then performed on the reference bit vector within the block to generate a value corresponding to the local detection block. Candidate list Based on the local corrected observations obtained in step (4) For the candidate list Local metrics are calculated for each candidate. Obtain the corresponding local list of fine-tuning software information. ;
[0013] (6) Detail the soft information based on the local list. Compared with the initial soft information obtained from the first stage of global initial detection Perform a fusion update to obtain the final detection output. .
[0014] As a preferred technical solution, in step (1), the first stage of global initial detection includes, but is not limited to, detection based on Gaussian approximation, detection based on minimum mean square error, detection based on sum-product algorithm, detection based on projection gradient, detection based on Bayesian Gaussian approximation, and other detection methods that can generate symbol-level or bit-level initial soft information.
[0015] As a preferred technical solution, in step (2), the reliability of each location is calculated based on the initial soft information output in the first stage. Specifically:
[0016] First, locations with low reliability after the initial global check are selected from the entire DD domain grid, and subsequent local list fine-checks are only performed on these locations. Since each symbol location in high-order modulation typically corresponds to multiple bits, the initial soft information, including symbol-level initial metrics, is used. Or the initial log-likelihood ratio ;definition ,in The constellation symbol with the largest initial symbolic metric. The constellation symbol with the second largest initial metric at the symbol level; or defined .
[0017] As a preferred technical solution, in step (3), the coupling relationship is constructed based on the coupling weight in the time-delay-Doppler domain equivalent channel matrix, and the local blocks are formed by connected component partitioning, threshold clustering, greedy expansion or restricted block growth, and the size of each local detection block is constrained by a preset maximum block size.
[0018] As a preferred technical solution, in step (4), the local correction observation... Constructed in any one or more of the following ways:
[0019] Soft sign estimation at off-block locations is used to cancel off-block soft interference;
[0020] Off-block interference can be canceled by using hard decision results from off-block locations.
[0021] As a preferred technical solution, in step (5), the benchmark decision is determined based on the initial detection information obtained in step (1), specifically including any of the following methods: based on the symbol-level initial metric. Select the candidate symbol with the best metric as the benchmark symbol decision; based on the initial log-likelihood ratio. Hard decision is performed on each bit to obtain a reference bit vector; constellation mapping is performed based on the reference bit vector to obtain a reference symbol vector; or, the hard decision result directly given by the first-stage global initial detection is used as the reference decision.
[0022] As a preferred technical solution, in step (5), the candidate list is generated by controlling the perturbation of the local detection block around the benchmark decision; the candidate list is generated in the bit flip domain or the sign flip domain, and adopts any one or more of the following methods: candidate generation method based on Hamming weight increment principle, candidate generation method based on soft weight sorting, candidate generation method based on integer factorization, and a hybrid candidate generation method formed by combining at least two of the above methods.
[0023] As a preferred technical solution, in step (5), the local metric evaluation includes one or a combination of the following: evaluation based on Euclidean distance, evaluation based on likelihood function, evaluation based on maximum logarithmic approximation, evaluation based on weighted cost function, or joint evaluation that integrates prior information outside the block and observation information inside the block.
[0024] As a preferred technical solution, in step (6), the soft information fusion and update includes one of the following: linear weighted fusion, convex combination fusion, selective replacement based on confidence comparison, updating only the position to be re-examined and retaining the first stage output for the remaining positions.
[0025] As a preferred technical solution, in step (6), the final detection output is the symbol-level detection result, bit-level soft output or bit-level log-likelihood ratio corresponding to each delay-Doppler position, and is used for subsequent channel decoding processing.
[0026] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0027] (1) The present invention adopts a two-stage processing framework of global initial detection and local list fine detection, and performs selective fine detection only on low reliability positions, which can effectively control the average computational complexity.
[0028] (2) By dividing the blocks according to the coupling relationship and correcting the observation outside the blocks, the present invention enables the local list fine inspection to focus on the key positions with significant interference, which is conducive to improving the accuracy of local fine inspection.
[0029] (3) The present invention allows different types of global soft detection methods to be used in the first stage and different candidate list generation rules to be used in the second stage, which has good versatility and scalability.
[0030] (4) By fusing and updating the global initial soft information and the local fine detection soft output, the present invention can improve the final detection performance while maintaining overall stability. Attached Figure Description
[0031] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0032] Figure 1 This is a schematic diagram of the OTFS two-stage signal detection method based on global initial detection and local list fine detection in an embodiment of the present invention;
[0033] Figure 2 In this embodiment of the invention, the transmitting end uses 5G LDPC code. Re-detection threshold under BPSK modulation Constructing thresholds with graphs A diagram illustrating the impact on the complexity of the second stage;
[0034] Figure 3 In this embodiment of the invention, the transmitting end uses 5G LDPC code. A schematic diagram comparing the performance of the global initial detection output based solely on Gaussian approximation under BPSK modulation with that of the fine detection output after flipping according to the Hamming weight increment principle in the second stage.
[0035] Figure 4 In this embodiment of the invention, the transmitting end uses 5G LDPC code. Re-detection threshold under QPSK modulation Constructing thresholds with graphs A diagram illustrating the impact on the complexity of the second stage;
[0036] Figure 5 In this embodiment of the invention, the transmitting end uses 5G LDPC code. A schematic diagram comparing the performance of the global initial detection output using only Gaussian approximation under QPSK modulation with that of the fine detection in the second stage, which is flipped according to the Hamming weight increment principle.
[0037] Figure 6 In this embodiment of the invention, the transmitting end uses 5G LDPC code. A schematic diagram comparing the performance of the global initial detection output based solely on Gaussian approximation under QPSK modulation with that of the fine detection output in the second stage, which is flipped according to the soft weight principle. Detailed Implementation
[0038] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative effort are within the scope of protection of the present application.
[0039] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.
[0040] Example 1
[0041] This embodiment provides a two-stage OTFS signal detection method based on global initial detection and local list fine detection, including the following steps:
[0042] S1, Receive signal according to OTFS and the corresponding time-delay-Doppler domain equivalent channel matrix The first stage of global initial detection is performed to obtain the initial detection information corresponding to each time delay-Doppler position, including the symbol-level initial metric corresponding to each DD domain position. This also includes the initial log-likelihood ratio of each bit within each position. .
[0043] Specifically, the time-delay-Doppler domain receiver matrix obtained after OTFS demodulation at the receiver is vectorized to obtain the received signal. And establish the input-output relationship at the receiving end. .in Send symbol vectors for the DD domain (Delay-Doppler Domain). For the additive noise vector in the DD domain to satisfy .
[0044] Based on global observation information across the entire delay-Doppler domain receive grid, global soft detection is performed at each delay-Doppler location to obtain initial detection information for each location. This initial detection information includes, but is not limited to, symbol-level initial metrics corresponding to each DD domain location. The initial log-likelihood ratio of each bit within each position .
[0045] Furthermore, in one embodiment, the first-stage global initial detection can employ a detection method based on Gaussian approximation; in other embodiments, other global soft detection methods capable of outputting initial soft information can also be employed, including but not limited to minimum mean square error detection, sum-product algorithm detection, projection gradient detection, and Bayesian Gaussian approximation detection.
[0046] S2. Determine the set of locations to be re-detected based on the initial soft information output from the first stage. .
[0047] Specifically, for each time-delay-Doppler position, the initial log-likelihood ratio Calculate the reliability at each location. And determine the set of locations to be re-inspected based on reliability. When the reliability of a certain location is lower than the preset re-detection threshold. When this happens, the location will be included in the set of locations to be re-inspected. .
[0048] Further, based on the initial soft information output from the first stage, the reliability of each location is calculated. Specifically:
[0049] First, locations with low reliability after the initial global check are selected from the entire DD domain grid, and subsequent local list fine-checks are only performed on these locations. Since each symbol location in high-order modulation typically corresponds to multiple bits, the initial soft information, including symbol-level initial metrics, is used. Or the initial log-likelihood ratio .definition ,in The constellation symbol with the largest initial symbolic metric. The constellation symbol with the second largest initial metric at the symbol level; or defined .
[0050] S3. Divide the data into local blocks based on the coupling relationship between the locations to be re-inspected.
[0051] Specifically, the preferred method is based on the coupling weights between different positions in the time-delay-Doppler domain equivalent channel matrix. Construct an association graph and a set of locations to be re-examined. Perform connected component partitioning to form one or more local detection blocks. .
[0052] Furthermore, the coupling relationship is constructed based on the coupling weights in the time-delay-Doppler domain equivalent channel matrix. The local blocks are formed by connected component partitioning, threshold clustering, greedy expansion, or restricted block growth, and the size of each local detection block is constrained by a preset maximum block size.
[0053] Furthermore, when the size of a certain connected component exceeds the preset maximum block size... The connected component can be further divided using a greedy expansion rule or a restricted block growth rule to control the computational complexity of subsequent local list refinement.
[0054] S4. For each local detection block Construct local observation corrections .
[0055] Specifically, first determine the local detection block. The relevant observation set and the set of associated locations outside the block are then used for soft sign estimation of the locations outside the block. Alternatively, an out-of-block interference cancellation term can be constructed from the hard decision result, and this out-of-block interference cancellation term can be subtracted from the received observations to obtain the local fine-detection observations corresponding to that local detection block. .
[0056] Furthermore, the local correction observation Constructed in any one or more of the following ways:
[0057] Soft sign estimation at off-block locations is used to cancel off-block soft interference;
[0058] Hard interference outside the block is canceled by using the hard decision result of the location outside the block;
[0059] S5. Based on the initial detection information obtained from the first stage global initial detection, and around the benchmark decision, a local candidate list is generated through controlled flipping. Local measurement evaluation is performed on each candidate to obtain local list fine-check soft information.
[0060] Specifically, the benchmark decision is determined based on the initial detection information obtained in step (1), specifically including any of the following methods: based on the symbol-level initial metric. Select the candidate symbol with the best metric as the benchmark symbol decision; based on the initial log-likelihood ratio. Hard decision is performed on each bit to obtain a reference bit vector; constellation mapping is performed based on the reference bit vector to obtain a reference symbol vector; or, the hard decision result directly given by the first-stage global initial detection is used as the reference decision.
[0061] Furthermore, in one implementation, the local candidate list can be generated using a bit-flipping rule based on the Hamming weight increment principle; in other implementations, a hybrid candidate generation rule based on soft weight sorting, integer factorization, reliability sorting, sign flipping, or a combination of the foregoing methods can also be used.
[0062] Furthermore, for each candidate in the local candidate list, further refinement can be performed based on local detailed observations. Calculate the local Euclidean distance, likelihood function value, maximum logarithmic approximation measure, or other local cost function, and obtain local optimal candidate or local list refinement soft information based on these. .
[0063] Furthermore, the benchmark decision includes one of the following: the hard decision result of the first stage output, the optimal symbol candidate of the first stage output, the benchmark bit vector derived from the initial soft information of the first stage, and the benchmark symbol vector derived from the initial soft information of the first stage.
[0064] S6. Generate local soft output based on the local fine inspection results, and merge and update it with the initial soft information of the first stage to obtain the final detection output.
[0065] Specifically, it is preferable to update the soft information only at the locations to be re-detected, while directly retaining the first-stage output for locations not included in the set of locations to be re-detected; for the locations to be re-detected, the initial soft information of the first stage can be fused using a linear weighted or convex combination method. Soft output with local list fine-tuning The final detection output is obtained. Preferably, the fusion process can be represented as follows: ,in, This is the fusion coefficient.
[0066] In this embodiment, through the processing flow of global initial detection, selection of locations to be re-detected, coupling block, local list fine detection, and soft information fusion update, controlled candidate search is performed only on locations with low reliability and significant local coupling. Therefore, the detection accuracy of key locations can be improved while keeping the overall complexity controllable.
[0067] Example 2
[0068] This embodiment 2 provides a performance example of a two-stage OTFS signal detection method based on global initial detection and local list fine detection, illustrating the application effect of the method of the present invention under BPSK modulation conditions. Consider an coded OTFS system, wherein the transmitting end adopts a base-based... Figure 2 The 5G LDPC encoding has a code length of n=108, an information length of k=54, a code rate of R=1 / 2, and uses BPSK modulation. Since each modulation symbol carries one coded bit, one frame of OTFS transmission corresponds to MN=108 delay-Doppler domain symbol positions. Preferably, the number of Doppler positions is N=6, and the number of delay positions is M=18. The maximum delay index is set to 3, and the maximum Doppler index is set to 5.
[0069] In this embodiment, the first stage employs a global initial detection method based on Gaussian approximation to output initial soft information to the entire time-delay-Doppler domain receiving grid. The second stage employs a local list refinement method based on the Hamming weight increment principle, that is, around the benchmark decision output in the first stage, a candidate list is generated sequentially in ascending order of the number of flipped symbols, and local metrics are evaluated on each candidate. Preferably, the maximum allowable block size for the local detection block is S. max =27, the maximum flip order is taken as t. max =4, the fusion coefficient is taken as 4. 0.4.
[0070] Furthermore, to illustrate the threshold for re-examination Constructing thresholds with graphs Impact on the complexity of the second-stage local list refinement Figure 2 The average number of candidates per frame under different parameter combinations is given. The result varies with signal-to-noise ratio. (From...) Figure 2 It can be seen that, in a fixed When, increase It will significantly reduce The reason is that, With the increase in size, the edge conditions in the coupling graph become more stringent, and weaker coupling relationships between locations to be re-detected are no longer preserved. This results in smaller and more dispersed local detection blocks, ultimately reducing the number of candidate enumerations in each local block and lowering the average total number of candidates per frame. Furthermore, under a fixed... When, increase It will significantly increase This is because After the size is increased, more locations are included in the set to be re-examined. This increases the number of locations to be re-examined in the second stage, further increasing the size of the local block and the total number of candidate enumerations. Therefore, Figure 2 This demonstrates how setting a reasonable threshold for re-examination can help. Constructing thresholds with graphs This invention can effectively control the candidate size of the second-stage local list refinement, thereby achieving a trade-off between detection performance and computational complexity. Furthermore, by... Figure 3 It can be seen that, under the different path count conditions examined, the proposed two-stage detection method consistently outperforms the baseline detector that only uses global initial detection, achieving a gain of approximately 2 dB under these parameter configurations and selections. This indicates that, based on the soft information obtained from the first-stage global initial detection, further local list-based fine-tuning of low-reliability locations can effectively improve the reliability of soft information at critical locations, thereby improving subsequent decoding performance.
[0071] Example 3
[0072] This embodiment 3 provides a performance example of a two-stage OTFS signal detection method based on global initial detection and local list fine detection, illustrating the application effect of the method of the present invention under QPSK modulation conditions. Consider a coded OTFS system where the transmitting end uses a linear block code with a code length of n=512 and an information length of k=256, and a code rate of R=1 / 2. Since each modulation symbol in QPSK modulation carries 2 coded bits, one frame of OTFS transmission corresponds to MN=256 time-delay-Doppler domain symbol positions. Preferably, the number of Doppler positions is N=16, and the number of time-delay positions is M=16. The maximum time-delay index is set to 3, and the maximum Doppler index is set to 5.
[0073] In this embodiment, the first stage employs a global initial detection method based on Gaussian approximation to output initial soft information to the entire delay-Doppler domain receiving grid. The second stage employs a local list refinement method based on the Hamming weight increment principle, that is, around the benchmark decision output in the first stage, a candidate list is generated sequentially in ascending order of the number of flipped bits, and local metric evaluation is performed on each candidate. Preferably, the maximum allowable block size for the local detection block is S. max =8, the maximum bit flip order is t. b =6, the fusion coefficient is taken as 6. 0.4.
[0074] Furthermore, under the above parameter conditions, the two-stage detection method used in this embodiment can be compared with the global initial detection method that only uses the first stage based on Gaussian approximation. The results are as follows: Figure 4 This indicates that by reasonably setting the threshold for re-examination... Constructing thresholds with graphs This invention can effectively control the candidate size of the second-stage local list refinement, thereby achieving a trade-off between detection performance and computational complexity. Furthermore, by... Figure 5 It can be seen that, under different path number conditions, the proposed two-stage detection method is always superior to the baseline detector that only uses global initial detection, and achieves more gain under this parameter configuration and selection.
[0075] Example 4
[0076] This embodiment 4 provides a performance example of a two-stage OTFS signal detection method based on global initial detection and local list fine-tuning, illustrating the application effect of the method of the present invention under QPSK modulation conditions when using global initial detection based on Gaussian approximation and local list fine-tuning based on soft weight principle flipping. Consider a coded OTFS system where the transmitting end uses a linear block code with a code length of n=512 and an information length of k=256, and a code rate of R=1 / 2. Since each modulation symbol in QPSK modulation carries 2 coded bits, one frame of OTFS transmission corresponds to MN=256 time-delay-Doppler domain symbol positions. Preferably, the number of Doppler positions is N=16 and the number of time-delay positions is M=16. The maximum time-delay index is set to 3, and the maximum Doppler index is set to 5.
[0077] In this embodiment, the first stage also employs a global initial detection method based on Gaussian approximation to output initial soft information to the entire delay-Doppler domain receiving grid. Unlike Embodiment 3, the second stage employs a local list refinement method based on the soft weight principle. Specifically, based on the initial soft information output in the first stage, the reliability of each bit or symbol in the local detection block is sorted, and low-reliability positions are preferentially flipped, thereby forming a candidate list generated according to the soft weight principle. Local metric evaluation is then performed on each candidate. Preferably, the maximum allowable block size for the local detection block is S. max =8, the maximum bit flip order is t. b =6, the fusion coefficient is taken as 6. 0.4.
[0078] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0079] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.
Claims
1. A two-stage signal detection method for OTFS based on global initial detection and local list fine detection, characterized in that, Includes the following steps: (1) Based on the OTFS received signal and the corresponding time-delay-Doppler domain equivalent channel matrix The first stage of global initial detection is performed to obtain initial detection information corresponding to each time delay-Doppler position. The initial detection information includes symbol-level initial metrics corresponding to each DD domain position. This also includes the initial log-likelihood ratio of each bit within each position. ; (2) Calculate the reliability of each location based on the initial soft information output from the first stage obtained in step (1). And determine the set of locations to be re-inspected based on reliability. , where the set Includes components with reliability below a preset threshold The time delay-Doppler position; (3) Based on the set of locations to be re-inspected obtained in step (2) The coupling relationships between different locations are calculated, and the coupling weights between locations are determined. And construct the threshold according to the coupling weight and graph. The comparison results are for the set of locations to be re-detected. Perform local segmentation to form one or more local detection blocks. ; (4) For each local detection block obtained in step (3) Utilizing initial detection information from relevant locations outside the block Computational soft symbol estimation And based on soft sign estimation Correcting local observations yields the corresponding local detection block. Local correction observations ; (5) For the local detection block obtained in step (3) Based on the initial log-likelihood ratio obtained in step (1) Hard decision is performed on each bit within the local detection block to obtain a reference bit vector within the block; controlled bit flipping is then performed on the reference bit vector within the block to generate a value corresponding to the local detection block. Candidate list Based on the local correction observations obtained in step (4) For the candidate list Local metrics are calculated for each candidate. Then by local measurement Obtain the corresponding local list of detailed inspection information. ; (6) Detail the soft information based on the local list. Compared with the initial soft information obtained from the first stage of global initial detection Perform a fusion update to obtain the final detection output. .
2. The two-stage OTFS signal detection method based on global initial detection and local list fine detection according to claim 1, characterized in that, In step (1), the first stage of global initial detection includes, but is not limited to, detection based on Gaussian approximation, detection based on minimum mean square error, detection based on sum-product algorithm, detection based on projection gradient, detection based on Bayesian Gaussian approximation, and other detection methods that can generate symbol-level or bit-level initial soft information.
3. The two-stage OTFS signal detection method based on global initial detection and local list fine detection according to claim 1, characterized in that, In step (2), the reliability of each location is calculated based on the initial soft information output in the first stage, specifically as follows: First, locations with low reliability after the initial global check are selected from the entire DD domain grid, and subsequent local list fine-checks are only performed on these locations. Since each symbol location in high-order modulation typically corresponds to multiple bits, the initial soft information, including symbol-level initial metrics, is used. Or the initial log-likelihood ratio ;definition ,in The constellation symbol with the largest initial symbolic metric. The constellation symbol with the second largest initial metric at the symbol level; or defined .
4. The two-stage OTFS signal detection method based on global initial detection and local list fine detection according to claim 1, characterized in that, In step (3), the coupling relationship is constructed based on the coupling weight in the time-delay-Doppler domain equivalent channel matrix. The local blocks are formed by connected component partitioning, threshold clustering, greedy expansion or restricted block growth, and the size of each local detection block is constrained by a preset maximum block size.
5. The two-stage OTFS signal detection method based on global initial detection and local list fine detection according to claim 1, characterized in that, In step (4), the local correction observation... Constructed in any one or more of the following ways: Soft sign estimation at off-block locations is used to cancel off-block soft interference; Hard interference outside the block is canceled by using the hard decision result of the location outside the block.
6. The two-stage OTFS signal detection method based on global initial detection and local list fine detection according to claim 1, characterized in that, In step (5), the benchmark decision is determined based on the initial detection information obtained in step (1), specifically including any of the following methods: based on the symbol-level initial metric. Select the candidate symbol with the best metric as the benchmark symbol decision; based on the initial log-likelihood ratio. Hard decision is performed on each bit to obtain a reference bit vector; constellation mapping is performed based on the reference bit vector to obtain a reference symbol vector; or, the hard decision result directly given by the first-stage global initial detection is used as the reference decision.
7. The two-stage OTFS signal detection method based on global initial detection and local list fine detection according to claim 1, characterized in that, In step (5), the candidate list is generated by controlling the perturbation of the local detection block around the benchmark decision; the candidate list is generated in the bit flip domain or the sign flip domain, and adopts any one or more of the following methods: candidate generation method based on Hamming weight increment principle, candidate generation method based on soft weight sorting, candidate generation method based on integer factorization, and hybrid candidate generation method formed by combining at least two of the above methods.
8. The two-stage OTFS signal detection method based on global initial detection and local list fine detection according to claim 1, characterized in that, In step (5), the local metric evaluation includes one or a combination of the following: evaluation based on Euclidean distance, evaluation based on likelihood function, evaluation based on maximum logarithmic approximation, evaluation based on weighted cost function, or joint evaluation that integrates prior information outside the block and observation information inside the block.
9. The two-stage OTFS signal detection method based on global initial detection and local list fine detection according to claim 1, characterized in that, In step (6), the soft information fusion and update includes one of the following: linear weighted fusion, convex combination fusion, selective replacement based on confidence comparison, updating only the position to be re-inspected and retaining the first stage output for the remaining positions.
10. The two-stage OTFS signal detection method based on global initial detection and local list fine detection according to claim 1, characterized in that, In step (6), the final detection output is the symbol-level detection result, bit-level soft output, or bit-level log-likelihood ratio corresponding to each delay-Doppler position, and is used for subsequent channel decoding processing.