Otfs signal detection method based on selective message delivery

Abstract

The application relates to a kind of orthogonal time-frequency space (OTFS) signal detection methods based on message selective transmission, suitable for high-speed mobile communication scene.The application first adds the calculation of symbol log-likelihood ratio (LLR) on the basis of original MP detection algorithm, improves the accuracy of message transmission, and optimizes the factor graph structure by introducing edge grouping strategy to group edges with different channel coefficients, reduces the influence of loop effect on detection performance.Then, a damping factor is introduced in the iteration process to control the message oscillation amplitude in each iteration and accelerate the convergence speed of the algorithm.During the detection process, the D-MsP detector further improves the detection performance and convergence efficiency of the OTFS signal in the multipath channel by using the sparsity principle.The D-MsP detector achieves a good balance between performance and computational complexity, significantly reduces the computational overhead while ensuring high detection accuracy, and is particularly suitable for high-speed mobile communication scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, and mainly relates to an OTFS signal detection method based on selective message transmission. BACKGROUND

[0002] With the rapid development of wireless communication technology, the demand for efficient and stable signal detection methods is increasing. In a multipath and high-speed mobile environment, the traditional orthogonal frequency division multiplexing (OFDM) system will have performance bottlenecks due to the influence of frequency selective fading. In order to solve this problem, orthogonal time frequency space modulation (OTFS) has emerged. OTFS uniformly distributes information symbols in the time-frequency domain, so that the robustness of the transmission signal in the multipath environment is significantly improved. However, with the wide application of OTFS technology, the complexity of signal detection has also increased significantly. In the OTFS system, the channel matrix is complex and contains a large number of path gains and noise interference. Traditional detection methods, such as linear minimum mean square error (LMMSE) detector, can estimate the prior probability of transmission symbols by optimizing the weight, reduce the computational complexity and improve the error performance. However, the LMMSE detector only performs well under certain conditions, and its detection performance will decrease when the channel complexity increases. Therefore, how to further improve the accuracy and convergence speed of signal detection has become the focus of research.

[0003] In recent years, message passing algorithm based on factor graph has been widely used in signal detection in wireless communication. Message passing algorithm realizes efficient symbol detection by iteratively transmitting information on the factor graph. However, in the OTFS system, the sparsity characteristics of the channel matrix make it difficult to directly use traditional message passing algorithm, and may cause information oscillation problem.

[0004] The present application proposes an OTFS signal detection method based on selective message transmission. This method combines the initial estimation ability of LMMSE detector and the iterative optimization characteristics of factor graph message passing algorithm. Through the grouping strategy of path gain, the message passing path is divided into "selected edge set" and "simplified edge set", and the symbol messages in different sets are treated differently. In addition, a damping factor is introduced in the iteration process to further stabilize the message transmission. This method realizes efficient and stable OTFS signal detection, and is suitable for communication systems in multipath propagation complex and high-speed mobile scenarios. SUMMARY

[0005] The present application aims to solve the problems of the above prior art. An OTFS signal detection method based on selective message transmission is proposed. The technical scheme of the present application is as follows:

[0006] An OTFS signal detection method based on message selective transmission, comprising the following steps:

[0007] Step 1, initialize the prior probability of the transmitted symbol according to the received signal and the channel matrix using the LMMSE detector, and obtain the initialized symbol prior information;

[0008] Step 2, calculate the initial message through the log-likelihood ratio (LLR) formula. The initial conditional probability of the symbol provides a basis for subsequent iterations.

[0009] Step 3, divide the transmitted symbol related to each received symbol into a "selected edge set" and a "simplified edge set" based on path gain;

[0010] Step 4, construct a sparse factor graph, and perform message selective transmission iteration based on the sparse factor graph;

[0011] Step 5, introduce a damping factor in each iteration to control the oscillation amplitude of message transmission;

[0012] Step 6, repeat the iteration until the maximum number of iterations is reached or the convergence condition is met, and output the final LLR result of the symbol.

[0013] 2. The OTFS signal detection method based on message selective transmission according to claim 1, wherein the step 1 is specifically: using the received signal y, the channel matrix H, and the noise power σ 2 Constructing a transmission symbol vector is:

[0014]

[0015] wherein, represents an estimated value, is a transmission symbol vector estimation vector, and H is an NM×NM dimensional equivalent channel matrix in the DD domain. Let the indices of the equivalent channel matrix row and column be d and c, respectively, that is, H={h dc ,d∈[1,MN],c∈[1,MN]},and I d and J c represent the index set of the position of the non-zero element in the dth row and the cth column of the matrix, respectively, and |I d |=|J c |=P, that is, there are only P non-zero elements in each row and each column of the matrix H, and the subscript MMSE represents the abbreviation of Minimum Mean Square Error (MMSE), and P is the number of multipaths. H represents a conjugate transpose operation, (·) -1 represents an inverse operation, I represents an MN×MN dimensional unit matrix, M is the number of subcarriers, and N is the number of symbols.

[0016] According to The initial prior probability of symbol a k is:

[0017]

[0018] where p j (a k ), j∈{1,2,...,MN} is the probability of the jth symbol being a k , represents the kth modulation symbol in the constellation, is the constellation base, exp(·) is the exponential function, and ||·||2 2 represents the square of the Euclidean distance.

[0019] 3. The OTFS signal detection method based on message selective transmission according to claim 1, wherein the step 2 is specifically: the log-likelihood ratio message from the symbol node to the function node f yi is calculated according to the initial prior probability in the step 1, and is

[0020]

[0021] where (·) (l) indicates the lth iteration, is the log-likelihood ratio message transmitted from the symbol node to the function node , and log(·) is the logarithmic function, p (l) (x j =a k |y, H) is the probability of the symbol x j being estimated as a k under the given received signal y and channel matrix H in the lth iteration, and p (l) (x j =a1|y, H) is the probability of the symbol x j being estimated as a1 under the given received signal y and channel matrix H in the lth iteration.

[0022] 4. The OTFS signal detection method based on message selective transmission according to claim 1, wherein the step 3 is specifically: the relevant transmitted symbols of the received symbols are divided into two sets according to the path gain. Specifically, the non-zero elements of the relevant rows in the channel matrix are sorted, and the symbols corresponding to the indexes are divided into a selected edge set and a simplified edge set. The symbol messages in the selected edge set remain unchanged, and the symbol messages in the simplified edge set only retain the maximum value.

[0023] 5. The OTFS signal detection method based on selective message passing according to claim 1, characterized in that step 4 specifically comprises: establishing a sparse factor graph based on the set partitioning in step 3, and performing selective message passing iteration on this basis, wherein the function node passes messages to the symbol node as follows:

[0024]

[0025] For the k-th modulation symbol from the function node To symbol node The transmitted log-likelihood ratio message, max(·) is the objective function to maximize. For the reason The resulting 1×P dimensional row vector For the reason The resulting P×1 dimensional column vector The constraint condition represents the transmitted symbol x. j Must be a k , Let be the set of indices containing the non-zero elements in column t. Symbol node To function node The log-likelihood ratio message passed during the (l-1)th iteration. For the k-th modulation symbol from the function node To symbol node The log-likelihood ratio message passed during the (l-1)th iteration.

[0026] Symbol nodes pass messages to function nodes as follows:

[0027]

[0028] Where J(j) is the index set of the non-zero elements in the j-th column.

[0029] 6. The OTFS signal detection method based on message selective delivery according to claim 1, wherein step 5 specifically comprises: introducing a damping factor Δ in each iteration of step 4. μ Specifically:

[0030]

[0031] Where Δ μ ∈[0,1) is a damping factor that improves performance by controlling the convergence rate.

[0032] 7. The OTFS signal detection method based on selective message passing according to claim 1, wherein step 6 specifically comprises: determining whether the maximum number of iterations I has been reached.max If satisfied, stop iteration and calculate the LLR of the symbol:

[0033]

[0034] where γ j (k) is the symbol node for a k total message amount.

[0035] Given a set of symbol LLRs, the factor graph of function nodes and related bit nodes c q (q∈{1,2,…,Q}) can derive its associated encoding bits, Q modulation order. Finally, using the sum-product algorithm and maximum log approximation, the LLR of the encoding bits can be calculated as:

[0036]

[0037] where, is the calculation result of the LLR of the qth bit encoding, is the maximum value of the encoding bit state being 1, is the maximum value of the encoding bit state being 0.

[0038] The advantages and beneficial effects of the present application are as follows:

[0039] The present application provides an OTFS signal detection method based on message selective transmission, aiming to improve the detection performance of OTFS signal in complex multipath and high-speed mobile environment. The method first uses the LMMSE detector to preliminarily estimate the received signal, and reduces the calculation complexity and improves the detection accuracy by initializing the prior probability of the symbol. Then, based on the difference of symbol path gain, the symbol is divided into selective edge set and simplified edge set, and the selective message passing algorithm is implemented on the factor graph to ensure efficient detection of symbol message under complex channel conditions.

[0040] The present method controls the oscillation amplitude of message passing through damping factor, improves the stability and convergence speed of the algorithm. Specifically, in each iteration, the symbol message in the simplified edge set is optimized to retain the maximum value, thereby reducing redundant calculation and improving system efficiency. At the same time, when the maximum iteration number is reached or the preset convergence condition is met, the log-likelihood ratio (LLR) of the symbol is output as the final detection result, ensuring the reliability and accuracy of the detection.

[0041] By combining LMMSE detection and factor graph message passing, the application realizes efficient demodulation of the received signal, and reduces inter-symbol interference through a symbol grouping strategy of path gain. The method performs excellently in a complex multipath propagation environment, and is suitable for 5G / 6G mobile communication, Internet of Vehicles, and unmanned aerial vehicle communication, and other application scenarios that require high reliability and low latency.

[0042] Through the above technical design, the application not only improves the detection accuracy of the OTFS signal, but also reduces the required computing resources. The introduction of path gain division and damping factors effectively controls the instability and oscillation problems in the computing process. In different channel environments, the method can ensure the stability and efficiency of detection, and provides strong support for high-speed mobile communication and application under complex channel conditions. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 is a OTFS system block diagram for a high-speed mobile scenario provided by the application;

[0044] Figure 2 is a flow chart of a low-complexity OTFS signal detection method for a high-speed mobile scenario provided by the application;

[0045] Figure 3 is a factor graph model diagram of MP detection provided by the application; DETAILED DESCRIPTION

[0046] In order to make the purpose, technical scheme and advantages of the application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application.

[0047] The application considers a single-input single-output orthogonal time-frequency space system, as shown in Figure 1 OTFS maps data streams to the DD domain for data transmission, and the OTFS system block diagram is as shown in Figure 1 The symbol in the DD domain is used, Q is the modulation order of the system, and the data symbol of the symbol letter constructs the OTFS signal, wherein k [0, N-1] and l [0, M-1] represent the Doppler shift index and the time delay index respectively.

[0048] The symbol x[k,l] is first converted into a TF domain symbol X[n,m] using inverse-sine finite Fourier transform (ISFFT), which is

[0049]

[0050] wherein n=0,...,N-1, m=0,...,M-1 are time index and frequency index respectively.

[0051] Then, the TF domain symbol X[n,m] is transformed into a continuous-time signal s(t) by Heisenberg Transform, denoted as:

[0052]

[0053] where g tx (t) denotes the transmit pulse shaping, Δf and denote the subcarrier spacing and symbol duration, respectively.

[0054] The time domain signal s(t) is then transmitted through a doubly selective channel, and the received signal can be written as

[0055] r(t) = ∫∫h(τ, v)e j2πν(t-τ) s(t - τ)dτdν + n(t)

[0056] where ∫∫(·)dτdv denotes the double integration of a function, h(τ, v) is the complex channel impulse response, representing the channel response to the pulse with Doppler v and delay τ. The sparse representation of the channel h(τ, v) in the DD domain satisfies

[0057]

[0058] where P is the number of propagation paths, h i , τ i , and v i are the channel gain, delay, and Doppler shift associated with the i-th path, respectively, and δ(·) is the impulse response function.

[0059] At the receiver, we use g rx (t) as the receive pulse, and transform the received signal r(t) into the TF domain by Wigner Transform, denoted as

[0060]

[0061] Y[n,m] = Y(t,f) t＝nT,f＝mΔf

[0062] The input-output relationship in the TF domain can be simplified by using ideal shaping

[0063] Y[n,m] = H[n,m]X[n,m] + V[n,m]

[0064] where V[n,m] is an additive white Gaussian noise (AWGN) with zero mean and variance σ 2 , and the equivalent channel H[n,m] in the TF domain is

[0065] H[n,m] = ∫∫h(τ,ν)e j2πνnT e -j2π(ν+mΔf)τ dνdτ

[0066] Finally, the TF signal Y[n,m] is mapped back to the DD domain received signal using the Sine Finite Fourier Transform (SFFT), denoted as

[0067]

[0068] Using the above equation, the input-output relationship in the DD domain can be derived as

[0069]

[0070] where is the sampled form of the circular convolution of the channel response and the window function w(τ,v).

[0071] Further, we stack y[k,l] into a vector y, i.e.:

[0072] y = Hx + n

[0073] where is the received vector, x k+Nl = x[k,l] is the transmitted symbol vector, is the AWGN vector, denotes the space of MNx1 complex matrices, and H is the NM=NM dimensional equivalent channel matrix in the DD domain. Let the indices of the rows and columns of the equivalent channel matrix be d and c, respectively, i.e. H = {h dc ,d∈[1,MN],c∈[1,MN]}, and I d and J c denote the index sets of the positions of the non-zero elements in the d-th row and c-th column of the matrix, respectively, and |I d | = |J c | = P, i.e. there are only P non-zero elements in each row and column of the matrix H.

[0074] The principles and implementation procedures of the method are described in detail as follows. Specifically, as shown in Figure 2

[0075] The OTFS signal detection method based on message selective transmission has the characteristics that the method comprises the following steps:

[0076] 1. An OTFS signal detection method based on message selective transmission, characterized in that the method comprises the following steps:

[0077] ​Step 1, initialize the prior probability of the transmitted symbol according to the received signal and the channel matrix, using the LMMSE detector, to obtain the initialized symbol prior information;

[0078] Using the received signal y, the channel matrix H, and the noise power σ 2 Construct the transmitted symbol vector For:

[0079]

[0080] Wherein, represents the estimated value, is the transmitted symbol vector estimation vector, and H is an NM×NM dimensional equivalent channel matrix in the DD domain. Let the indexes of the equivalent channel matrix row and column be d and c, respectively, that is, H={h dc ,d∈[1,MN],c∈[1,MN]},and I d and J c respectively represent the index set of the position of the non-zero element in the dth row and the cth column of the matrix, and |I d |=|J c |=P, that is, there are only P non-zero elements in each row and each column of the matrix H, and the subscript MMSE represents the abbreviation of the English name of the minimum mean square error (Minimum Mean Square Error), and P is the number of multipath.(·) H represents the conjugate transpose operation, (·) -1 represents the inverse operation, I represents an MN×MN dimensional unit matrix, M is the number of subcarriers, and N is the number of symbols.

[0081] According to the initial prior probability of the symbol a k is calculated as:

[0082]

[0083] Wherein, p j (a k ), j∈{1,2,...,MN} is the probability that the jth symbol is a k , represents the kth modulation symbol in the constellation, is the base of the constellation, exp(·) is the exponential function, and ||·|| 2 represents the square of the Euclidean distance.

[0084] Step 2, calculate the initial message through the log-likelihood ratio (LLR) formula. The initial conditional probability of the symbol provides a basis for subsequent iterations.

[0085] According to the initial prior probability in step 1, the initial message is calculated from the symbol node to the function node the log-likelihood ratio message for

[0086]

[0087] where (·) (l) denotes the l-th iteration, the log-likelihood ratio message passed from the function node to the symbol node , log(·) is the logarithm function, p (l) (x j = a k | y, H) is the probability that the symbol x j is a k given the received signal y and the channel matrix H at the l-th iteration. (l) (x j = a1| y, H) is the probability that the symbol x j is a1given the received signal y and the channel matrix H at the l-th iteration.

[0088] Step 3, divide the transmitting symbols related to each received symbol into a "selection edge set" and a "simplified edge set" based on path gains;

[0089] According to the path gains, the transmitting symbols related to the received symbol are divided into two sets. Specifically, the non-zero elements of the related rows in the channel matrix are sorted, and the symbols corresponding to the indexes are divided into a selection edge set and a simplified edge set. The symbol messages in the selection edge set remain unchanged, and the symbol messages in the simplified edge set only retain the maximum value.

[0090] Step 4, construct a sparse factor graph, and perform message selective transmission iteration based on the sparse factor graph;

[0091] The sparse factor graph is established according to the set division in step 3, and the message selective transmission iteration is performed based on the sparse factor graph. The message passed from the function node to the symbol node is:

[0092]

[0093] the log-likelihood ratio message passed from the function node to the symbol node , max(·) is the maximization objective function, is a 1xP-dimensional row vector composed of , is a Px1-dimensional column vector composed of , is a constraint condition, indicating that the transmitting symbol x j must be a k , is the index set of the position of the non-zero element in the t-th column, is the symbol node is the function node is the log-likelihood ratio (LLR) message passed at the (l-1)-th iteration, is the function node is the symbol node is the log-likelihood ratio (LLR) message passed at the (l-1)-th iteration.

[0094] The symbol node passes the message to the function node as:

[0095]

[0096] where J(j) is the index set of the position of the non-zero element in the j-th column.

[0097] Step 5. Introduce a damping factor in each round of iteration to control the oscillation amplitude of message passing;

[0098] Introduce a damping factor Δ μ in each round of iteration in step 4, specifically:

[0099]

[0100] where Δ μ ∈ [0, 1) is a damping factor to improve performance by controlling the convergence speed.

[0101] Step 6. Repeat the iteration until the maximum number of iterations is reached or the convergence condition is met, and output the final LLR result of the symbol.

[0102] Determine whether the maximum number of iterations I max is reached, if so, stop the iteration, and calculate the LLR of the symbol:

[0103]

[0104] γ j (k) is the total message amount of the symbol node a k .

[0105] Given a set of symbol LLRs, the associated encoding bits can be derived from the factor graph of the function node and the related bit nodes c q (q ∈ {1, 2,..., Q}), where Q is the modulation order. Finally, using the sum-product algorithm and maximum log approximation, the LLR of the encoding bits can be calculated as:

[0106]

[0107] where, LLR calculation result for the qth position, maximum value for the state of the encoded bit being 1, maximum value for the state of the encoded bit being 0.

[0108] As described above, the application can be better implemented. Other contents not described in detail in the specification of the application belong to the known technology of the skilled in the art, and will not be specifically described.

[0109] The following is a low-complexity OTFS signal detection method for high-speed mobile scenarios using the application described above, and the convergence and performance of the algorithm in the prior art are compared. Through comparison, it can be seen that compared with the prior art, the application not only has advantages in convergence speed, but also has good performance in high-speed mobile scenarios.

[0110] The above examples should be understood as only for illustrating the application and not for limiting the protection scope of the application. After reading the content described in the application, the skilled in the art can make various modifications or modifications to the application, and these equivalent changes and modifications also fall within the scope defined by the claims of the application.

Claims

1. An OTFS signal detection method based on message selective delivery, characterized in that, The method includes the following steps: Step 1: Based on the received signal and the channel matrix, initialize the prior probability of the transmitted symbol using an LMMSE detector to obtain the initialized symbol prior information; using the received signal y, the channel matrix H, and the noise power σ... 2 Construct the transmission symbol vector x MMSE for: x MMSE =(H H H+σ 2 I) -1 H H y Where (·) represents the estimated value, x MMSE Let H be the transmission symbol vector estimation vector, and H be the NM×NM dimensional equivalent channel matrix in the DD domain; let the row and column indices of the equivalent channel matrix be d and c, respectively, i.e., H = {h dc ,d∈[1,MN],c∈[1,MN]}, and use I d and J c Let |I| represent the index sets of the non-zero elements in the d-th row and c-th column of the matrix, respectively, and |I| d |=|J c | = P, meaning that each row and column of matrix H has only P non-zero elements. The subscript MMSE stands for Minimum Mean Square Error, and P is the number of multipaths; (·) H This indicates the conjugate transpose operation, (·). -1 This indicates the inversion operation, where I represents the MN×MN dimensional identity matrix, M is the number of subcarriers, and N is the number of symbols; According to x MMSE , calculation symbol a k The initial prior probability is: Where, p j (a k ), j∈{1,2,...,MN} is the j-th symbol a k The probability, This represents the k-th modulation symbol in the constellation. Let exp(·) be the cardinality of the constellation, and ||·|| be the exponential function. 2 Represents the square of the Euclidean distance; Step 2: Calculate the initial message using the log-likelihood ratio (LLR) formula. The initial conditional probability based on the symbol provides the basis for subsequent iterations; the initial prior probability is calculated from the symbol node. To function node The log-likelihood ratio of the message is in(·) (l) Refers to the l-th iteration. For the k-th modulation symbol from the symbol node To function node The transmitted log-likelihood ratio message, log(·) is the logarithmic function, p (l) (x j =a k |y,H) is the symbol x j Given the received signal y and the channel matrix H, it is estimated as a in the l-th iteration. k The probability, p (l) (x j =a1|y,H) is the symbol x j Given the received signal y and the channel matrix H, the probability that it is estimated as a1 in the l-th iteration; Step 3: Based on path gain, the transmitted symbols associated with each received symbol are divided into a "selection edge set" and a "simplified edge set"; by sorting the non-zero elements of the relevant rows in the channel matrix, the symbols corresponding to the indices are divided into a selection edge set and a simplified edge set; the symbol messages in the selection edge set remain unchanged, while the symbol messages in the simplified edge set retain only the maximum value; Step 4: Construct a sparse factor graph, and perform selective message passing iterations based on it; the message passing from the function node to the symbol node is as follows: in, For the k-th modulation symbol from the function node To symbol node The transmitted log-likelihood ratio message, max(·) is the objective function to maximize, h i For the reason A 1×P dimensional row vector is formed, where x is a vector composed of... The P×1 dimensional column vector, x:x j =a k The constraint condition represents the transmitted symbol x. j Must be Let be the set of indices containing the non-zero elements in column t. Symbol node To function node The log-likelihood ratio message passed during the (l-1)th iteration. For the k-th modulation symbol from the function node To symbol node The log-likelihood ratio message passed during the (l-1)th iteration; Symbol nodes pass messages to function nodes as follows: Where J(j) is the index set of the non-zero elements in the j-th column; Step 5: Introduce a damping factor Δ in each iteration. μ This is to control the oscillation amplitude of message transmission; specifically: Where Δ μ ∈[0,1) is a damping factor that improves performance by controlling the convergence rate; Step 6: Repeat the iterations until the maximum number of iterations is reached or the convergence condition is met, and output the final LLR result with the symbol; determine whether the maximum number of iterations I has been reached. max If satisfied, stop iterating and compute the LLR of the symbol: γ j (k) is a symbol node For a k Total message volume; Given a set of LLRs with symbols, based on function nodes and related bit nodes c q The factor graph of (q∈{1,2,...,Q}) can be used to derive its associated coded bits and Q modulation order; finally, using the sum-product algorithm and the maximum logarithm approximation, the LLR of the coded bits is calculated as follows: in, The LLR calculation result for the q-th bit encoding. The maximum value of the encoded bit state being 1. This represents the maximum value of the encoded bit state being 0.

Images