OTFS system signal detection method based on Turbo structure
Through the iterative information transmission method of the Turbo structure, combined with the message passing algorithm, MMSE detector and LDPC decoder, the problem of unbalanced performance and complexity in the OTFS system is solved, the detection performance is improved and the system complexity is reduced.
Patent Information
- Application Number
- CN202510838630.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-23
- Publication Date
- 2025-09-12
AI Technical Summary
The existing OTFS signal detection method significantly increases the algorithm complexity when improving the detection performance, making it difficult to balance performance and complexity.
A signal detection method based on the Turbo structure is adopted. Through the iterative process between the message passing algorithm and the MMSE detector, combined with the LDPC decoder, the information transmission between the detector and the decoder is utilized to reduce the complexity and improve the detection performance.
The detection performance of the OTFS system is improved at a lower complexity, and the reliability and detection effect of the system are enhanced.
Smart Images
Figure CN120639246A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of OTFS system signal detection methods, and in particular to an OTFS system signal detection method based on a Turbo structure. Background Art
[0002] Under the dual-channel time-frequency selection conditions of high-speed mobile scenarios, the widely used Orthogonal Frequency Division Multiplexing (OFDM) modulation suffers from severe Doppler shift, which destroys the orthogonality of its subcarriers, severely degrading receiver performance and making it unsuitable for high-speed environments. Orthogonal Time-Frequency-Space (OTFS) modulation has been proposed as a new modulation scheme. OTFS multiplexes information symbols in the Delay-Doppler (DD) domain rather than the Time-Frequency (TF) domain. OTFS uses a two-dimensional transformation to fully expand each symbol in the TF domain, resulting in a similar channel effect on each symbol, resulting in significant performance gains in highly dynamic communications.
[0003] However, this also poses new challenges to the OTFS system's detectors. Balancing the complexity and performance of the OTFS system's detection algorithm becomes a difficult problem. Many researchers have developed various low-complexity detectors by exploiting the sparsity of wireless channels in the DD domain. Summary of the Invention
[0004] The present invention aims to provide an OTFS system signal detection method based on a Turbo structure, aiming to solve the technical problem that in order to obtain better detection performance, the existing OTFS signal detection method causes a significant increase in algorithm complexity.
[0005] To achieve the above object, the OTFS system signal detection method based on the Turbo structure includes the following steps:
[0006] Step 1: Perform message passing algorithm detection on the channel output signal that has been demodulated by OTFS. The output mean and variance are used as the initial values for MMSE detection in the first iteration of step 3.
[0007] Step 2: Determine whether the current number of iterations has reached the preset maximum number. If so, exit the iteration and output the final detection result. Otherwise, proceed to step 3.
[0008] Step 3: Perform MMSE detection on the demodulated OTFS signal. The mean and variance of the first MMSE detection are provided by the message passing detection algorithm. The subsequent mean and variance are calculated by calculating the extrinsic information output by the LDPC. The MMSE detector calculates the extrinsic information of each symbol and feeds it into the LDPC channel decoder.
[0009] Step 4: The LDPC decoder decodes the information input by the MMSE detector to obtain new external information;
[0010] Step 5: Calculate the mean and variance of the new external information as the input prior information of the MMSE detector in the next iteration;
[0011] Step 6: Return to step 2 and enter the next round of iteration, with the number of iterations + 1;
[0012] Alternatively, for signals that have been demodulated by OTFS, the message passing detector is modeled as a sparsely connected factor graph with NM variable nodes corresponding to x and NM variable nodes corresponding to y. In this factor graph, each observation node y[d] is associated with a set of S variable nodes Similarly, each observation node x[c] is connected to a set of S variable nodes For each y[d], the variable x[c] is isolated from other disturbances, which are approximated as Gaussian noise with easy-to-compute mean and variance.
[0013] (1) The observation node y[d] is passed to the variable node The calculation formula is as follows:
[0014] Mean of the distractors and variance is approximately modeled as a Gaussian random variable It is defined as follows:
[0015]
[0016] Gaussian random variable The mean for:
[0017]
[0018] variance for:
[0019]
[0020] (2) The variable node x[c] is passed to the observation node The calculation formula is as follows:
[0021] Probability mass function pmf vector Updated to:
[0022]
[0023] Where Δ∈(0,1] is the damping factor that improves the performance by controlling the convergence speed.
[0024]
[0025] where ξ (i) (e, c, k) is the probability density function of the normal distribution, and the calculation formula is as follows:
[0026]
[0027] (3) Convergence index η (i) :
[0028]
[0029] For some small γ>0 where
[0030]
[0031] is an indicator function that evaluates to 1 if the argument expression is true, and 0 otherwise.
[0032] (4) Update of Judgment:
[0033] If η (i) >η (i-1) , then we update the decision of the transmission symbol to:
[0034]
[0035] Stopping criteria:
[0036] The MP algorithm can stop when at least one of the following conditions is met.
[0037] a.η (i) =1.
[0038] b.η (i) <η (i*) -∈ where i * ∈0,…i-1 is η (i*) The iteration index of the maximum value.
[0039] c. Reach the maximum number of iterations n iter .
[0040] (5) Calculation of mean and variance
[0041]
[0042] where the vector The receiving vector and the noise vector have elements y[d] and z[d] respectively; is the delay-Doppler channel matrix, whose elements are H[d,c]; is the information symbol vector, whose elements are x[c]∈A, where d,c∈[0,...,NM-1],A is the QAM modulation symbol set. denotes the set of indices with non-zero elements in row d and column c, respectively, and for all rows and columns
[0043] Optionally, the estimation result of the MMSE detector is calculated as follows
[0044]
[0045] Among them, ξ n , V, A calculation formula is as follows
[0046]
[0047] V=diag[v1,...,v MN ]
[0048]
[0049] The output external information calculation formula is as follows
[0050]
[0051] Among them, h n Indicates H DD The nth column, m=[m1,...,m MN ] T , represents the mean vector, The result of the estimated symbol, m n and v n is the mean and variance of the symbol, N0 is the mean of the Gaussian white noise channel, I is the unit matrix, L e (d n ,j),j=1,2,is the external LLR value output by the MMSE detector.
[0052] Optionally, the LDPC decoding is calculated as follows:
[0053] (1) Initialization: For variable node i, its connected check node j∈C i Initialization information sent
[0054] L (0) (q ij )=L(P i )=L e (d n ,j)
[0055] (2) Information update of the check node: In the lth iteration, the check node updates the information of the variable node i∈Rj connected to it
[0056]
[0057] β is the offset factor, satisfying β>0. The optimal setting of β is also related to multiple factors, including the structural parameters of the LDPC code, channel conditions, and decoding iteration strategy. It is usually set to a constant, generally in the range of 0.1-0.5.
[0058] (3) Update variable node information
[0059] In a new round of iteration, the variable node i is connected to the check node j∈C i The information delivered is updated.
[0060]
[0061] (4) Decoding decision
[0062] Calculate decoding decision information:
[0063]
[0064] If L (l) (q i )>0, then otherwise
[0065] (5) Determine whether decoding is terminated:
[0066]
[0067] If the above equation holds true or the set maximum number of iterations is reached, LDPC decoding ends.
[0068] (6) Output new external information L′ e (C n ,j),j=1,2。
[0069] L′ e (C n ,j)=L (1) (q i )-L (0) (q ij )
[0070] Optional, for L e ′(C n ,j) Calculate its mean and variance, the calculation formula is as follows:
[0071]
[0072] L e ′(C n ,j),j=1,2, represents the external information LLR output by the LDPC encoder BRIEF DESCRIPTION OF THE DRAWINGS
[0073] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0074] Figure 1 This is a flow chart of a Turbo-structure-based OTFS system signal detection method of the present invention.
[0075] Figure 2 It is the BER data of the OTFS system signal detection algorithm of the present invention and the commonly used MMSE detection algorithm
[0076] Figure 3 It is a block diagram of the OTFS system.
[0077] Figure 4 is the exchange of messages in the MP algorithm factor graph DETAILED DESCRIPTION
[0078] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.
[0079] See also Figure 1 , a Turbo structure-based OTFS system signal detection method, comprising the following steps:
[0080] S1: Perform message passing algorithm detection on the channel output signal that has been demodulated by OTFS. The output mean and variance are used as the initial values for MMSE detection in the first iteration of step 3.
[0081] S2: Determine whether the current number of iterations has reached the preset maximum number. If so, exit the iteration and output the final detection result. Otherwise, proceed to step 3.
[0082] S3: Perform MMSE detection on the demodulated OTFS signal. The mean and variance of the first MMSE detection are provided by the message passing detection algorithm. The subsequent mean and variance are calculated by calculating the extrinsic information output by the LDPC. The MMSE detector calculates the extrinsic information of each symbol and feeds it into the LDPC channel decoder.
[0083] S4: The LDPC decoder decodes the information input by the MMSE detector to obtain new external information;
[0084] S5: Calculate the mean and variance of the new external information as the input prior information of the MMSE detector in the next iteration;
[0085] S6: Return to step 2 and enter the next round of iteration, the number of iterations + 1;
[0086] Specifically, existing signal detection algorithms for Orthogonal Time Frequency and Space (OTFS) modulation systems cannot effectively balance performance and complexity. Furthermore, the signal detection and channel decoding processes are independent of each other. After signal detection, the results are output to the decoder for decoding, and there is no iterative information transfer between the two. To address these issues, a signal detection method for OTFS systems based on a Turbo architecture is proposed. This algorithm utilizes the iterative structure of Turbo and continuously transfers information between the detector and decoder, fully utilizing the effective information of both. Furthermore, the MMSE detection algorithm and the MP algorithm used in this algorithm result in a low-complexity LDPC encoder, achieving good detection performance while ensuring that the complexity is not excessively high.
[0087] Further, the following is an explanation in conjunction with specific implementation steps:
[0088] Step S1: For the signal that has been demodulated by OTFS, the message passing detector is modeled as a sparsely connected factor graph with NM variable nodes corresponding to x and NM variable nodes corresponding to y. In this factor graph, each observation node y[d] is associated with a set of S variable nodes Similarly, each observation node x[c] is connected to a set of S variable nodes For each y[d], the variable x[c] is isolated from other disturbances, which are approximated as Gaussian noise with easy-to-compute mean and variance.
[0089] (1) The observation node y[d] is passed to the variable node The calculation formula is as follows:
[0090] Mean of the distractors and variance is approximately modeled as a Gaussian random variable It is defined as follows:
[0091]
[0092] Gaussian random variable The mean for:
[0093]
[0094] variance for:
[0095]
[0096] (2) The variable node x[c] is passed to the observation node The calculation formula is as follows:
[0097] Probability mass function pmf vector Updated to:
[0098]
[0099] Where Δ∈(0,1] is the damping factor that improves the performance by controlling the convergence speed.
[0100]
[0101] where ξ (i) (e, c, k) is the probability density function of the normal distribution, and the calculation formula is as follows:
[0102]
[0103] (3) Convergence index η (i) :
[0104]
[0105] For some small γ>0 where
[0106]
[0107] is an indicator function that evaluates to 1 if the argument expression is true, and 0 otherwise.
[0108] (4) Update of Judgment:
[0109] If η (i) >η (i-1) , then we update the decision of the transmission symbol to:
[0110]
[0111] Stopping criteria:
[0112] The MP algorithm can stop when at least one of the following conditions is met.
[0113] d.η (i) =1.
[0114] e. where i * ∈0,…i-1 is The iteration index of the maximum value.
[0115] f. Reach the maximum number of iterations n iter .
[0116] (5) Calculation of mean and variance
[0117]
[0118] where the vector The receiving vector and the noise vector have elements y[d] and z[d] respectively; is the delay-Doppler channel matrix, whose elements are H[d,c]; is the information symbol vector, whose elements are x[c]∈A, where d,c∈[0,...,NM-1],A is the QAM modulation symbol set. denotes the set of indices with non-zero elements in row d and column c, respectively, and for all rows and columns
[0119] Step S3: pass the mean and variance into the MMSE detector. The calculation formula is as follows:
[0120]
[0121] Among them, ξ n , V, A calculation formula is as follows
[0122]
[0123] V=diag[v1,...,v MN ]
[0124]
[0125] The output external information calculation formula is as follows
[0126]
[0127] Among them, h n Indicates H DDThe nth column, m=[m1,...,m MN ] T , represents the mean vector, The result of the estimated symbol, m n and v n is the mean and variance of the symbol, N0 is the mean of the Gaussian white noise channel, I is the unit matrix, L e (d n ,j),j=1,2,is the external LLR value output by MMSE detector
[0128] Step S4: Send the MMSE detection result to the LDPC decoder. The LDPC decoding calculation is as follows:
[0129] (1) Initialization: For variable node i, its connected check node j∈C i Initial
[0130] Initialization information
[0131] L (0) (q ij )=L(P i )=L e (d n ,j)
[0132] (2) Information update of the check node: In the lth iteration, the check node updates the information of the variable node i∈Rj connected to it
[0133]
[0134] β is the offset factor, satisfying β>0. The optimal setting of β is also related to multiple factors, including the structural parameters of the LDPC code, channel conditions, and decoding iteration strategy. It is usually set to a constant, generally in the range of 0.1-0.5.
[0135] (3) Update variable node information
[0136] In a new round of iteration, the variable node i is connected to the check node j∈C i The information delivered is updated.
[0137]
[0138] (4) Decoding decision
[0139] Calculate decoding decision information:
[0140]
[0141] If L (l) (q i )>0, then otherwise
[0142] (5) Determine whether decoding is terminated:
[0143]
[0144] If the above equation holds true or the set maximum number of iterations is reached, LDPC decoding ends.
[0145] (6) Output new external information L′ e (C n ,j),j=1,2。
[0146] L′ e (C n ,j)=L (1) (q i )-L (0) (q ij )
[0147] To L e ′(C n ,j) Calculate its mean and variance, the calculation formula is as follows:
[0148]
[0149] L e ′(C n ,j),j=1,2,represents the extrinsic information LLR output by the LDPC encoder.
[0150] The new mean and variance are sent to the MMSE detector for the next iteration until the maximum number of iterations is reached. The detection and decoding process ends and the final result is output.
[0151] Using a turbo iterative structure, multiple iterations between the detector and decoder fully utilize the information between the decoder and detector, gradually improving signal detection performance. Before iterative detection, MP detection is performed as preprocessing, further improving performance. The MMSE and MP algorithms are relatively simple, effectively ensuring the overall complexity of the detector.
[0152] pass Figure 2 From the data, it can be seen that the OTFS system signal detection method based on Turbo structure proposed in this paper shows excellent detection performance. Figure 2This paper clearly demonstrates that the proposed signal detection algorithm improves bit error rate compared to both the commonly used MMSE algorithm without channel coding and the MMSE detection algorithm with LDPC coding (when iter is 1). Performance improves significantly when iter is 2, but increases further. This is because the mutual information available to the signal detector and channel decoder decreases, so significant improvements typically require no more than three iterations. This innovative technology improves the detection performance and enhances the reliability of the OTFS communication system by leveraging effective information from signal detection and channel decoding.
[0153] The above describes the specific embodiments of the present invention in detail, but it should be understood that the present invention is not limited to the above specific embodiments. Those skilled in the art may make various modifications or variations without exceeding the scope of the claims of the present invention, without affecting the core essence of the present invention. In the absence of conflict, the various embodiments and features in this application may also be combined in any manner.
Claims
1. A method for detecting signals in an OTFS system based on a Turbo structure, comprising the following steps: Step 1: Perform message passing algorithm detection on the channel output signal that has been demodulated by OTFS. The output mean and variance are used as the initial values of MMSE detection in the first iteration. Step 2: Determine whether the current number of iterations has reached the preset maximum number. If so, exit the iteration and output the final detection result. Otherwise, proceed to step 3. Step 3: Perform MMSE detection on the demodulated OTFS signal. The mean and variance of the first MMSE detection are provided by the message passing detection algorithm. The subsequent mean and variance are calculated by calculating the extrinsic information output by the LDPC. The MMSE detector calculates the extrinsic information of each symbol and feeds it into the LDPC channel decoder. Step 4: The LDPC decoder decodes the information input by the MMSE detector to obtain new external information; Step 5: Calculate the mean and variance of the new external information as the input prior information of the MMSE detector in the next iteration; Step 6: Return to step 2 and enter the next round of iteration, with the number of iterations + 1.
2. The OTFS system signal detection method based on Turbo structure according to claim 1, characterized in that: For the signal that has been demodulated by OTFS, the message passing detector is modeled as a sparsely connected factor graph with NM variable nodes corresponding to x and NM variable nodes corresponding to y. In this factor graph, each observation node y[d] is associated with a set of S variable nodes. Similarly, each observation node x[c] is connected to a set of S variable nodes For each y[d], the variable x[c] is isolated from other disturbances, which are approximated as Gaussian noise with easy-to-compute mean and variance. The observation node y[d] is passed to the variable node The calculation formula is as follows: Mean of the distractors and variance is approximately modeled as a Gaussian random variable It is defined as follows: Gaussian random variable The mean for: variance for: The variable node x[c] is passed to the observation node The calculation formula is as follows: Probability mass function pmf vector Updated to: Where Δ∈(0,1] is the damping factor that improves the performance by controlling the convergence speed. where ξ (i) (e, c, k) is the probability density function of the normal distribution, and the calculation formula is as follows: Convergence index η (i) : For some small γ>0 where is an indicator function that evaluates to 1 if the argument expression is true, and 0 otherwise. Verdict Update: If η (i) >η (i-1) , then we update the decision of the transmission symbol to: Stopping criteria: The MP algorithm can stop when at least one of the following conditions is met. or (i) =1. where i * ∈0,…i-1 is The iteration index of the maximum value. Reach the maximum number of iterations n iter . Calculation of mean and variance: where the vector The receiving vector and the noise vector have elements y[d] and z[d] respectively; is the delay-Doppler channel matrix, whose elements are H[d,c]; is the information symbol vector, whose elements are x[c]∈A, where d,c∈[0,...,NM-1],A is the QAM modulation symbol set. denotes the set of indices with non-zero elements in row d and column c, respectively, and for all rows and columns 3. The OTFS system signal detection method based on Turbo structure according to claim 2, characterized in that: The detection result of the MMSE detector is calculated as follows: Among them, ξ n , V, A calculation formula is as follows: V=diag[v1,...,v MN ] The output external information calculation formula is as follows: Among them, h n Indicates H DD The nth column, m=[m1,...,m MN ] T , represents the mean vector, Indicates the result of the detected symbol, m n and v n is the mean and variance of the symbol, N0 is the mean of the Gaussian white noise channel, I is the unit matrix, L e (d n ,j),j=1,2,is the external LLR value output by the MMSE detector.
4. The OTFS system signal detection method based on Turbo structure according to claim 2, characterized in that: The LDPC decoding calculation is as follows: Initialization: For variable node i, check node j∈C connected to it i Initialization information sent L (0) (q ij )=L(P i )=L e (d n ,j) Information update of the check node: The lth iteration check node updates the variable node i∈R connected to it. j Information Update: β is the offset factor, satisfying β>0. The optimal setting of β is also related to multiple factors, including the structural parameters of the LDPC code, channel conditions, and decoding iteration strategy. It is usually set to a constant, generally in the range of 0.1-0.
5. Update variable node information: In a new round of iteration, the variable node i is connected to the check node j∈C i The information delivered is updated. Decoding decision: Calculate decoding decision information: If L (l) (q i )>0, then otherwise Determine whether decoding is terminated: If the above equation holds true or the set maximum number of iterations is reached, LDPC decoding ends. Output new external information L′ e (C n ,j),j=1,2。 L′ e (C n ,j)=L (1) (q i )-L (0) (q ij )。 5. The OTFS system signal detection method based on Turbo structure according to claim 2, characterized in that: To L e ′(C n ,j) Calculate its mean and variance, the calculation formula is as follows: L e ′(C n ,j),j=1,2,represents the extrinsic information LLR output by the LDPC encoder.