Signal detection method for OTFS system

Abstract

The invention discloses a signal detection method for an OTFS system, and the method comprises the steps: reducing the inter-symbol interference of the OTFS system in a multipath environment through placing a zero symbol in a time delay-time domain, deducing the power attenuation of different time delay transmission paths according to a V2V link loss model in a 3GPP standard, effectively simulating an actual V2V communication scene, and improving the detection precision of the OTFS system. A transmission path is divided into a detection path and an interference path through the effective signal to interference plus noise ratio of the path, so that the calculation complexity of time delay-time domain MRC detection is effectively reduced, and the performance and the calculation complexity are ensured to be well balanced. A multi-tap time delay-time domain MMSE equalizer is adopted to effectively reduce the number of iterations of time delay-time domain MRC detection, and a conjugate gradient method is used to replace matrix inversion in MMSE, so that the calculation complexity is reduced to O (KNM (2alpha-1)), the reliability of an OTFS system in a V2V scene is effectively improved, the calculation complexity of a detection algorithm is reduced, and the low-delay requirement is met.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technologies, and in particular, to a signal detection method for an OTFS system. Background Art

[0002] High-speed mobile scenarios are one of the important scenarios for future wireless communication networks, such as high-speed rail communication and vehicle-to-everything (V2X) communication. In some ultra-reliable and low-latency communication (URLLC) scenarios, the connection latency needs to reach the 1-ms level, and high-reliability connections in the case of high-speed movement (500 km / h) need to be supported.

[0003] Taking V2V communication as an example, due to the high-speed relative movement characteristics of vehicles, the channel will change rapidly, resulting in a doubly selective characteristic. Vehicles and other objects such as other vehicles and roadside units will cause the generation of multiple transmission paths, leading to frequency-selective fading. High-speed vehicle movement will generate a large Doppler frequency shift and expansion, resulting in significant inter-carrier interference (ICI). The large Doppler frequency shift will cause the loss of orthogonality between carriers in traditional OFDM systems, thus seriously degrading the communication performance of traditional OFDM systems.

[0004] To solve the problem of performance degradation in high-speed mobile scenarios or when there is a large amount of Doppler frequency offset in the channel, orthogonal time frequency space (OTFS) modulation is proposed as a new two-dimensional modulation technology. The OTFS technology can convert the time-frequency doubly selective channel into an approximately time-invariant stationary channel in the delay-Doppler domain through a two-dimensional transformation, reducing the impact of time-selective fading and frequency-selective fading. Compared with OFDM systems, it has better reliability. The equivalent DD (Delay-Doppler, DD) domain channel has excellent separability, stability, simplicity, and potential sparsity, enabling the OTFS system to obtain delay-Doppler diversity gain, thereby enhancing the ability of OTFS signals to resist Doppler frequency shift in high-speed mobile environments and improving V2X communication performance. Therefore, OTFS modulation is generally considered a potential solution to support broadband and reliable communication requirements in high-mobility scenarios.

[0005] In the paper "Interference cancellation and iterative detection for orthogonal time frequency space modulation" published by RAVITEJAP, HONG Y et al. (IEEE Transactions on Wireless Communications, 2018), a signal detection method for OTFS systems based on message passing (MP) is mentioned. This method establishes a sparse factor graph between the transmitted symbols and the received symbols according to the sparse characteristics of the equivalent channel matrix in the time-delay Doppler domain. The mean and variance of the interference terms are passed to the variable nodes through the observation nodes, and the probability mass function of the transmitted symbol set of the variable nodes is sent to the observation nodes. Such regular iterations are performed multiple times to complete the estimation of the transmitted symbols. To reduce the computational complexity, the interference terms are approximated as Gaussian white noise. The drawback of this method is that the Gaussian interference according to the central limit theorem is inaccurate, resulting in a decline in the performance of the signal detector of the message passing algorithm in the sparse channel. At the same time, when the number of signal paths is relatively large or there is a fractional Doppler frequency offset, the sparsity of the equivalent channel matrix will weaken, and the computational complexity of the message passing detection algorithm will increase sharply. Summary of the Invention

[0006] Aiming at the problem of high complexity of the signal detection algorithm in the OTFS system in the high-dynamic multi-transmission path scenario, the present invention proposes a signal detection method for the OTFS system.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] A signal detection method for the OTFS system, including:

[0009] Step 1: Perform constellation modulation on the transmitted signal from the information source and map it to the time-delay Doppler domain with M rows and N columns, and set the last l max rows of symbols to zero; where l max is the maximum time delay number of the path;

[0010] Step 2: Perform an N-point inverse fast Fourier transform on the rows of the data in the time-delay Doppler domain to transform it to the time-delay time domain, and serialize the symbols in the time-delay time domain by column and send them as discrete-time signals;

[0011] Step 3: Analyze the power attenuation of different time-delay transmission paths according to the 3GPP standard, construct a time-delay Doppler domain channel model and a time-delay time domain channel model, and obtain the input-output relationship in the time-delay time domain;

[0012] Step 4: Divide the transmission paths into detection paths and interference paths according to the path effective signal-to-interference-plus-noise ratio;

[0013] Step 5: Process the detection path signals in the time-delay - time domain. Use a multi-tap time-delay - time domain block decomposition least mean square error equalizer for initial estimation, and replace matrix inversion with the conjugate gradient method. The result is used as the initial value of maximum ratio combining detection;

[0014] Step 6: According to the input-output relationship in the time-delay - time domain and the initial value of maximum ratio combining detection, use the maximum ratio combining detection algorithm to recover the estimated value of the transmitted signal, and perform a decision to determine whether to continue iterative update or demodulate and output the transmitted signal.

[0015] Further, in the said Step 1, the frame structure in the time-delay - Doppler domain is based on a two-dimensional plane Γ with dimensions M×N:

[0016]

[0017] where l and k are integer indices in the time-delay and Doppler dimensions respectively, M and N are the numbers of subcarriers and symbols respectively, Δf = 1 / T and T are the adjacent subcarrier frequency interval and symbol period respectively, and l / MΔf and k / NT represent the quantization steps of delay and Doppler frequency respectively.

[0018] Further, in the said Step 3, the time-delay - Doppler domain channel model is:

[0019]

[0020] where, h DD (τ, ν) is the time-delay - Doppler domain channel impulse response function with respect to τ and ν, h i is the channel complex gain of the i-th transmission path, l i and κ i are the normalized time-delay offset and normalized Doppler frequency shift related to the i-th transmission path, P is the number of transmission paths, τ and ν are Doppler frequency shift variables, and δ(·) is the Dirac impulse function.

[0021] Further, in the said Step 3, the time-delay - time domain channel model is constructed in the following manner:

[0022] Based on the time-delay - Doppler domain channel model, perform an inverse Fourier transform on the Doppler frequency shift variable ν to obtain the continuous time-varying channel impulse response in the time-delay - time domain:

[0023]

[0024] After discretization, h DDSubstituting (τ, ν) and τ = l / MΔf into the above formula, we get:

[0025]

[0026] By sampling the waveform at the sampling interval t = qT / M, the time-delay - time domain channel is discretized through sampling, and a discrete baseband time-delay - time domain channel model is obtained.

[0027]

[0028] where 0 ≤ q ≤ NM - 1 and TΔf = 1. is the set of different time delays l in P paths in the time-delay - Doppler domain, and sinc represents the sinc function.

[0029] Furthermore, an integer time-delay tap is obtained. The time-delay - time domain channel model under the following is obtained:

[0030]

[0031] where L is the set of integer time-delay taps, and K l is the set of different Doppler frequency shifts κ under the same time delay.

[0032] Furthermore, step 4 includes:

[0033] Calculating the channel complex gain of all transmission paths according to the standard V2V link path loss model.

[0034] Then, based on the channel complex gains of different transmission paths, calculating the effective signal-to-interference-plus-noise ratio (SINR) of each path, and determining the detection paths and interference paths according to the effective SINR.

[0035] Furthermore, in step 4, the effective SINR of the path is calculated in the following manner:

[0036]

[0037] where E S represents the signal transmission power, h i is the channel complex gain of the i-th transmission path, N0 represents the noise power, represents the set of channel matrix coefficients that do not include h i , P is the number of transmission paths, and P′ represents the number of detection paths.

[0038] Furthermore, step 5 includes:

[0039] Dividing the two-dimensional data symbols in the time-delay - time domain at the receiving end into blocks in the time dimension to obtain and

[0040] Similarly, the two-dimensional data symbols in the time delay-time domain of the transmitting end are obtained.

[0041] The input-output relationship in the time delay-time domain of a single time block is obtained.

[0042]

[0043] where is the time delay-time domain channel matrix of a single time block, is the noise vector;

[0044] Substituting the input-output relationship into the minimum mean square error estimation formula, we get

[0045]

[0046] Regarding the above formula as a solution problem of the linear equation system Ax = b, where

[0047]

[0048] Solving the equation system by the conjugate gradient method, the time delay-time domain data symbols after initial estimation are obtained and used as the initial value of maximum ratio combining detection.

[0049] Furthermore, the estimated value of the transmitted signal is:

[0050]

[0051] In the formula,

[0052]

[0053] where is the received vector, The initial estimated value of is obtained from

[0054] Furthermore, the computational complexity of the multi-tap time delay-time domain block decomposition minimum mean square error equalizer is O(KNM(2α - 1)), where K is the condition number of the covariance matrix of is the time delay-time domain channel matrix of a single time block, α is the number of non-zero elements in each row of the covariance matrix, and α ≤ L, K and α are much smaller than NM.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] The present invention reduces the inter-symbol interference suffered by the OTFS system in a multipath environment by placing zero symbols in the delay-Doppler domain and the delay-time domain, and derives the power attenuation of the simultaneous delay transmission paths according to the V2V link loss model in the 3GPP standard to establish an actual channel model, effectively simulating the actual V2V channel scenario. The transmission paths are divided into detection paths and interference paths based on the path effective signal-to-interference-plus-noise ratio, and signal detection processing is only performed on the detection paths, effectively reducing the computational complexity of the delay-time domain MRC detection and ensuring a good balance between detection performance and computational complexity. By using a multi-tap delay-time domain minimum mean square error equalizer, the number of iterations of the delay-time domain MRC detection is effectively reduced, and the conjugate gradient method is used to replace the matrix inversion in MMSE, reducing the computational complexity to O(KNM(2α - 1)), and effectively reducing the inter-symbol interference caused by the multipath effect and improving the performance of the OTFS system in a multipath scenario. Then, the estimated value of the detection path signal is used as the initial value of the MRC detector, and the transmitted signal is recovered through the MRC detector. Since MRC is only performed on the detection paths, the MRC detector is further reduced due to the reduction of the combined paths. The computational complexity of the MRC detection proposed by the present invention is only O(n iter NML'), compared with the traditional MRC detector with a computational complexity of O(n iter NML 2 ), it is significantly reduced.

[0057] It can be seen from the experimental results that when the detection paths divided by the path effective signal-to-interference-plus-noise ratio of the present invention account for half of the total transmission paths, the best effect can be achieved, effectively reducing the computational complexity in a high-dynamic multi-transmission path scenario. At the same time, compared with other methods, the bit error rate is also significantly reduced. Description of the Drawings

[0058] Figure 1 It is a schematic flowchart of a signal detection method for an OTFS system according to an embodiment of the present invention;

[0059] Figure 2 It is an example of a V2V communication road model provided by an embodiment of the present invention;

[0060] Figure 3 It is a schematic diagram of two-dimensional data symbol division in the delay-time domain provided by an embodiment of the present invention;

[0061] Figure 4 It is a comparison of BER of different detection methods under 4-QAM modulation provided by an embodiment of the present invention;

[0062] Figure 5 It is a comparison of BER of different detection methods under 16-QAM modulation provided by an embodiment of the present invention;

[0063] Figure 6 Iteration times and bit error rate comparison of different solutions provided by embodiments of the present invention;

[0064] Figure 7 Iteration times and bit error rate comparison of different solutions provided by embodiments of the present invention;

[0065] Figure 8 Computational complexity comparison of different algorithms provided by embodiments of the present invention. Detailed implementation manners

[0066] The present invention will be further explained and described below in conjunction with the accompanying drawings and specific embodiments:

[0067] As Figure 1 shown, a signal detection method for an OTFS system includes:

[0068] Step 1: Perform constellation modulation on the transmitted signal from the information source, map it to the delay-Doppler domain of M rows and N columns, and set the last l max rows of symbols to zero; where l max is the maximum delay number of paths;

[0069] Step 2: Perform an N-point inverse fast Fourier transform (IFFT) on the rows of data in the delay-Doppler domain to transform it to the delay-time domain, and serialize the symbols in the delay-time domain column by column and send them as discrete-time signals;

[0070] Step 3: Analyze the power attenuation of different delay transmission paths according to the 3GPP standard, construct a delay-Doppler domain channel model and a delay-time domain channel model, and obtain the input-output relationship in the delay-time domain;

[0071] Step 4: Divide the transmission paths into detection paths and interference paths according to the path effective signal-to-interference-plus-noise ratio;

[0072] Step 5: In the receiving end, the detection path signals are detected and processed in the delay-time domain. Since the delay-time domain is not affected by the fractional Doppler frequency shift, a multi-tap delay-time domain block decomposition MMSE (Minimum Mean Squared Error) equalizer is used for initial estimation, and the conjugate gradient method is used to replace the matrix inversion, effectively reducing the computational complexity of the multi-tap delay-time domain block decomposition MMSE equalizer. The result is used as the initial value of the MRC detection (Maximal Ratio Combining);

[0073] Step 6: After initial estimation in the delay-time domain, the detection path signals use the MRC detection algorithm to recover the estimated value of the transmitted signal, and perform a decision to determine whether to continue iterative update or demodulate and output the transmitted signal.

[0074] Furthermore, in step 1, the frame structure of OTFS modulation in the delay-Doppler domain is based on a two-dimensional plane Γ with dimensions M×N, and this plane is denoted as:

[0075]

[0076] where l and k are integer indices in the delay and Doppler dimensions respectively, M and N are the number of subcarriers and the number of symbols respectively, Δf = 1 / T and T are the adjacent subcarrier frequency spacing and the symbol period respectively, and they are inversely proportional, Δf = 1 / T, where l / MΔf and k / NT represent the quantization steps of delay and Doppler frequency respectively. The information symbols sent from the source are mapped into the plane Γ after QAM modulation, and the last l max row symbols are set to zero. l max is the maximum number of delays of the path. The purpose of this operation is to prevent the multipath interference caused by the multipath effect. The symbol matrix in the delay-Doppler domain at the transmitter is obtained The corresponding symbol column vector can be expressed as x m = [X(m,0),...,X(m,N-1)], and the corresponding relationship between the two is X T = [x0,....,x M-1 . Similarly, the received symbol matrix in the delay-Doppler domain at the receiver can be written as and y m = [Y(m,0),...,Y(m,N-1)].

[0077] Furthermore, in step 2, the data in the delay-Doppler domain is subjected to an N-point inverse fast Fourier transform on the rows to transform it into the delay-time domain to obtain the information symbols in the delay-time domain The conversion method is and The symbols in the delay-time domain are subjected to serial-to-parallel conversion, and the transmission pulse uses a rectangular pulse shaping waveform G tx , which is equivalent to windowing in the time domain, and the information symbol vector in the discrete time domain of transmission can be obtained The conversion process can be written as Similarly, the received symbol matrix in the delay-time domain at the receiver can be written as and Let r be the information symbol vector in the discrete time domain at the receiver, To sum up, the relationship between the information symbols in the discrete time domain and the transmitted and received signals in the delay-time domain can be written as:

[0078]

[0079] Furthermore, in step 3, according to the 3GPP standard V2V link path loss model:

[0080] Line-of-sight (LOS): PL LOS = 32.4 + 20 log 10 (d 3D ) + 20 log 10 (f c )

[0081] Non-line-of-sight (NLOS): PL NLOS = 36.85 + 30 log 10 (d 3D ) + 18.9 log 10 (f c )

[0082] where d 3D represents the distance between the transmitter and the receiver in 3D space in meters. The path loss of the propagation path can be calculated from the speed and path delay of the object's movement, obtaining the relationship between the path delay and the path loss. f c is the carrier frequency in GHz.

[0083] Considering an equivalent channel model with P transmission paths, for a high-mobility wireless channel, the channel can be represented by a linear time-varying (LTV) system, i.e., a doubly selective channel. As an implementable manner, the present invention considers a V2V communication road as Figure 2 shown Figure 2 in which is a four-lane communication transmission model with two forward and two reverse lanes (i.e., P = 4), and there are direct and reflected links in the communication link. h i is the channel complex gain of the i-th transmission path, l i and κ i (l i , κ i (not necessarily integers) are the normalized delay offset and normalized Doppler shift associated with the i-th transmission path, and the channel impulse response function in the time-delay - Doppler domain can be obtained:

[0084]

[0085] It can also be written as

[0086]

[0087] where is the set of different delays l among the P paths in the time-delay - Doppler domain, κ ∈ K l , K l is the set of different Doppler shifts κ at the same delay.

[0088] Perform an IFFT (Inverse Fast Fourier Transform) on the Doppler shift variable ν to obtain a continuous time-varying channel impulse response that is continuous in the delay-time domain:

[0089]

[0090] After discretization, substituting h DD (τ, ν) and τ = l / MΔf into the above formula, we can obtain:

[0091]

[0092] By sampling the waveform at a sampling interval of t = qT / M, where 0 ≤ q ≤ NM - 1 and TΔf = 1, the delay-time channel is discretized through sampling to obtain a discrete baseband delay-time domain channel model

[0093]

[0094] where sinc represents the sinc function.

[0095] Assume that the channel delay offset is an integer, and the channel model can be simplified without loss of accuracy to obtain an integer-delay tap The delay-time domain channel model under:

[0096]

[0097] The received signal can be written as

[0098]

[0099] L is a set of integer-delay taps. Combining the information symbols in the discrete time domain and the relationship between the transmitted and received signals in the delay-time domain, let q = (m + nM), m = 0, 1, 2..., M - 1 and n = 0, 1, 2..., N - 1, and define We can obtain

[0100]

[0101] It can also be written as

[0102]

[0103] where is a banded matrix composed of and

[0104] Further, in step 4, the channel coefficients h of all transmission paths are calculated according to the standard V2V link path loss model in step 3, and the detection path and interference path are determined according to the effective signal-to-interference-plus-noise ratio

[0105]

[0106] where E S represents the signal transmission power, N0 represents the noise power, represents the set of channel matrix coefficients that does not include h i , P′ represents the number of detection paths, and the optimal effective signal-to-interference-plus-noise ratio is obtained by selecting different numbers of detection paths. Therefore, a suitable P′ can ensure a good balance between performance and computational complexity.

[0107] Further, in step 5, a multi-tap delay-time domain block decomposition MMSE (minimum mean square error) equalizer is used for initial estimation. The two-dimensional data symbols in the delay-time domain at the receiving end are divided into blocks in the time dimension to obtain as Figure 3 shown.

[0108] Similarly, the two-dimensional data symbols in the delay-time domain at the transmitting end can be obtained According to the literature (Thaj T, Viterbo E. Low complexity iterative rake decision feedback equalizer for zero-padded OTFS systems[J]. IEEE transactions on vehicular technology, 2020, 69(12):15606-15622)

[0109]

[0110] the input-output relationship in the delay-time domain of a single time block can be deduced

[0111]

[0112] where is the channel matrix in the delay-time domain of a single time block, is the noise vector. Substituting the input-output relationship into the minimum mean square error estimation formula, we can get

[0113]

[0114] By analyzing this formula, the above formula can be regarded as a problem of solving the linear equation system Ax = b, where

[0115]

[0116] Among them the result is a symmetric positive definite matrix, and the system of equations can be solved by the conjugate gradient method without using the traditional matrix inversion method, effectively reducing the computational complexity. The steps are as follows:

[0117] Step 1: Initialization

[0118] The initial solution is x0 = 0;

[0119] The initial residual

[0120] The initial search direction p0 = r0.

[0121] Step 2: Iterative update

[0122] Loop and execute the following steps until the convergence condition is met:

[0123] 1. Calculate the step size coefficient α k :

[0124]

[0125] 2. Update the estimated value x k+1 :

[0126] x k+1 = x k + α k p k

[0127] 3. Update the residual r k+1 :

[0128] r k+1 = r k - α k Ap k

[0129] 4. Calculate the conjugate coefficient β k :

[0130]

[0131] 5. Update the search direction p k+1 :

[0132] p k+1 = r k+1 + β k p k

[0133] Step 3: Termination condition

[0134] When ‖r kIteration terminates when 2 < ε, and x k is the approximate solution, that is

[0135] The time-delay time-domain data symbol after initial estimation can be written as

[0136]

[0137] The computational complexity of the multi-tap time-delay time-domain block decomposition MMSE (Minimum Mean Square Error) equalizer is O(KNM(2α - 1)), where K is the condition number of the covariance matrix α is the non-zero elements in each row of the covariance matrix, and α ≤ L, K and α are generally much smaller than NM. The computational complexity of the traditional minimum mean square error equalizer is O(N 3 M 3 ). The computational complexity of this scheme is much smaller than that of the traditional minimum mean square error equalizer. The multi-tap equalizer can more accurately compensate for each frequency component and time delay of the signal, thus more effectively solving problems such as multipath interference and frequency selective fading, and providing higher performance.

[0138] Furthermore, the time-delay time-domain MRC detection in step 6 can be understood as the maximum ratio combination of the channel-impaired components received by the two-dimensional data symbols in the time-delay time-domain with L ≤ P' different time-delay branches. The noise plus interference power in each branch is different and depends on the channel response. In each detector iteration, we cancel the estimated inter-symbol vector interference in the branches selected for combination, thereby iteratively improving the Signal-to-Interference-plus-Noise Ratio (SINR) of the signal after MRC.

[0139] According to the input-output relationship in the discrete time-delay time-domain in step 3, the relationship between each time-delay received block and the transmitted block that experiences different channel time delays can be deduced. It can be written as:

[0140]

[0141] where l ∈ L is the channel time delay of different detection paths, is the additive noise with variance σ 2 Due to the inter-symbol interference caused by the time-delay spread (l / MΔf), all symbol vectors in the received symbol vector will have a signal component, where l ∈ L. Assuming that the received vector after removing the interference of other transmitted symbols is the signal received vector We can conclude that and the symbol vector estimate value relation:

[0142]

[0143] where the symbol vector initial estimate value is obtained from the estimate value in step 5 , which can effectively reduce the number of iterations of the MRC detection algorithm. Combining we can conclude the relation of the received vector for different l ∈ L:

[0144]

[0145] When performing MRC detection in the delay-time domain, first define the residual noise plus interference term (RNPI) in the i-th iteration:

[0146]

[0147] To reduce the computational complexity of each iteration, the formula is changed to:

[0148]

[0149] In this way, only need to calculate once before MRC detection and only need to calculate in subsequent iterations to update to avoid repeated calculations thus reducing the overall computational complexity

[0150] It can be concluded that and the symbol vector estimate value relation in the i-th iteration:

[0151] The MRC detection formula for the i-th iteration is:

[0152]

[0153] where

[0154]

[0155] and

[0156] Substituting the above formula, the MRC detection calculation formula in the delay-time domain for the i-th iteration can be obtained:

[0157]

[0158] where

[0159] When the iteration stops, output After demodulating the transmitted signal is recovered.

[0160] The computational complexity of the traditional MRC method is O(n iter NML 2 ), and the computational complexity of the MRC scheme designed by the method of the present invention is O(n iter NML').

[0161] The beneficial effects of the present invention will be further elaborated below in combination with the simulation experiment results. The experimental tool for all simulation experiments is MATLAB.

[0162] Experiment 1: Compare the bit error rates of the method of the present invention with the existing MP method, traditional MRC method, and LU-MMSE method at a signal-to-noise ratio of 10 dB to 20 dB.

[0163] The simulation parameters are as follows:

[0164] Table 1 Simulation parameter settings for Experiment 1

[0165]

[0166] From Figure 4 it can be seen that at a speed of 120 km / h, compared with other OTFS detection methods, the bit error rate of the method of the present invention is significantly lower than that of other methods. Thanks to the divided detection paths, the low-complexity multi-tap delay-time domain MMSE equalizer designed by the present invention has a performance about 0.6 dB better than that of the traditional MRC and about 2.5 dB better than that of the existing MP method when the bit error rate is 10 -5 .

[0167] From Figure 5 it can be seen that the bit error rate of the scheme of the present invention is significantly reduced compared with the traditional MRC method and MP method under high SNR conditions. Since the detection path is only half of the total path, the scheme of the present invention is significantly lower than the traditional MRC and MP methods.

[0168] Experiment 2: Compare the multi-tap delay-time domain MMSE equalizer designed by the present invention with the single-tap equalizer of traditional MRC detection and the MRC scheme without initial estimation.

[0169] The simulation parameter settings are as follows:

[0170] Table 2 Simulation parameter settings for Experiment 2

[0171]

[0172]

[0173] From Figure 6 It can be seen that when SNR = 20, the present invention only needs about two iterations to reach convergence. Compared with the MRC detector without initial estimation, which requires 5 iterations to reach convergence, at the same time, the method of the present invention reduces the bit error rate performance by two orders of magnitude, verifying that the initial value provided by the multi-tap delay-time domain MMSE equalizer proposed by the present invention can effectively reduce the number of iterations of the MRC detector to quickly reach convergence and effectively reduce the bit error rate.

[0174] From Figure 7 It can be seen that when SNR = 20, both the proposed scheme of the present invention and the traditional single-tap MRC detector reach convergence in about two times. However, it can be seen from the figure that the bit error rate is significantly reduced, which can effectively reduce the inter-symbol interference caused by the multipath effect. And since the detection path is only half of the total path, the computational complexity is significantly lower than that of the traditional single-tap MRC detector. The experimental results verify that the present invention can effectively reduce the computational complexity, resist the inter-symbol interference caused by the multipath effect, and reduce the bit error rate in the high-dynamic multi-transmission path scenario.

[0175] Experiment 3: Verify that the method of the present invention can effectively reduce the computational complexity in the multipath scenario according to the computational complexity of different algorithms given in Table 3.

[0176] Simulation parameter settings:

[0177] Table 3 Simulation parameter settings for Experiment 3

[0178]

[0179]

[0180] Table 4 Computational complexity of different detection algorithms

[0181]

[0182] Table 4 shows the complexity of the comparison algorithms used, Figure 8 showing the computational complexity of different algorithms. The number of iterations of the MP algorithm is set to 15, and the number of iterations of the MRC algorithm and the algorithm of the present invention are Figure 6 given, and the maximum number of iterations is set to 10, and K is set to 4. According to the EVA channel model with a speed of 120 Km / h and the simulation parameters, it can be calculated that ∈ max = 16, τ max = 32. Through Figure 8 It can be concluded that the computational complexity of the traditional MP algorithm is relatively high. Although the LU-MMSE algorithm reduces the computational complexity from O((NM)3 ) reduced to O(NM[log2(M)+l 2 max +L 2 β]), however, in the scenario of high dynamic multi - transmission paths, the large number of paths and path delay are the main reasons for the increase in computational complexity. The computational complexity of the method of the present invention is significantly reduced compared with the MP algorithm and the LU - MMSE algorithm, and there is a certain reduction compared with the traditional MRC algorithm. Combining Figure 4 It can be seen that the method of the present invention significantly improves the communication performance under multi - transmission paths while reducing the computational complexity.

[0183] In summary, the technical key points and points to be protected by the present invention are as follows:

[0184] 1. A low - complexity multi - tap delay - time - domain MMSE equalizer. By decomposing the received - end delay - time - domain two - dimensional data symbols into symbol vectors in blocks, and analyzing that the covariance matrix in MMSE is a symmetric positive - definite matrix, the conjugate gradient method can be combined to replace the matrix inversion process in MMSE, reducing the computational complexity of the traditional minimum mean - square - error equalizer with O(N 3 M 3 ) to O(KNM(2α - 1)). Using the multi - tap delay - time - domain MMSE equalizer can compensate the various frequency components and delays of the signal more accurately compared with the single - tap time - frequency - domain MMSE equalizer, thus more effectively solving problems such as multi - path interference and frequency - selective fading, providing more accurate initial values to improve the reliability in the V2V scenario, and performing pre - estimation processing in the delay - time domain. Compared with the traditional time - frequency - domain equalizer, it can effectively avoid the extra calculations brought by the conversion between different domains.

[0185] 2. Compared with processing data in the delay - Doppler domain, the delay - time domain is not affected by fractional Doppler shift. Due to the characteristics of the delay - time domain, it can effectively separate the transmission paths under different delays, make full use of the delay diversity gain, and derive the power attenuation of different - delay transmission paths according to this characteristic combined with the V2V link loss model in the 3GPP standard, enabling the establishment of an actual channel model and effectively simulating the actual channel scenario. By dividing the transmission paths into detection paths and interference paths according to the path effective signal - to - interference - plus - noise ratio, and only performing signal detection processing on the detection paths, the computational complexity of delay - time - domain MRC detection in the V2V scenario is effectively reduced, and a good balance between performance and computational complexity is ensured.

[0186] The above - shown are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A signal detection method for an OTFS system, characterized in that, Including: Step 1: Perform constellation modulation on the transmitted signal from the source, map it to the time-delay Doppler domain of M rows and N columns, and set the last l max row symbols to zero; where l max is the maximum number of time delays of the path; Step 2: Perform an N-point inverse fast Fourier transform on the rows of the data in the delay-Doppler domain to transform it into the delay-time domain, serialize the symbols in the delay-time domain by column, and then send them as discrete-time signals. Step 3: Analyze the power attenuation of different delay transmission paths according to the 3GPP standard, construct a delay-Doppler domain channel model and a delay-time domain channel model, and obtain the input-output relationship in the delay-time domain. Step 4: Divide the transmission paths into detection paths and interference paths according to the path effective signal-to-interference-plus-noise ratio. Step 5: Process the detection path signals in the delay-time domain, use a multi-tap delay-time domain block decomposition least mean square error equalizer for initial estimation, and replace matrix inversion with the conjugate gradient method. The result is used as the initial value of the maximum ratio combining detection. Step 6: According to the input-output relationship in the delay-time domain and the initial value of the maximum ratio combining detection, use the maximum ratio combining detection algorithm to recover the estimated value of the transmitted signal, and perform a decision to determine whether to continue iterative update or demodulate and output the transmitted signal.

2. The signal detection method for an OTFS system according to claim 1, wherein In the said Step 1, the frame structure in the delay-Doppler domain is based on a two-dimensional plane Γ with dimensions M×N: where l and k are integer indices in the delay and Doppler dimensions respectively, M and N are the numbers of subcarriers and symbols respectively, Δf = 1 / T and T are the adjacent subcarrier frequency spacing and symbol period respectively, and l / MΔf and k / NT represent the quantization steps of delay and Doppler frequency respectively.

3. A signal detection method for an OTFS system according to claim 2, characterized in that In the said Step 3, the delay-Doppler domain channel model is: where h DD (τ, v) is the time-delay-Doppler domain channel impulse response function with respect to τ and v, and h i is the channel complex gain of the i-th transmission path, e i and k i are the normalized delay offset and normalized Doppler shift associated with the i-th transmission path, P is the number of transmission paths, τ and v are Doppler shift variables, and δ(·) is the Dirac impulse function.

4. A signal detection method for an OTFS system according to claim 3, characterized in that, In the said Step 3, construct the delay-time domain channel model in the following manner: Based on the delay-Doppler domain channel model, perform an inverse Fourier transform on the Doppler frequency shift variable ν to obtain the continuous time-varying channel impulse response in the delay-time domain: After discretization, substitute h DD (τ, ν) and τ = l / MΔf into the above formula to obtain: Sample the waveform at a sampling interval t = qT / M, discretize the delay-time domain channel through sampling, and obtain the discrete baseband delay-time domain channel model. where \(0\leq q\leq NM - 1\) and \(T\Delta f = 1\), is the set of different delays \(l\) in \(P\) paths in the time-delay Doppler domain, and sinc represents the sinc function; Thus, an integer-delay tap is obtained. The delay-time domain channel model under: where L is a set of integer time-delay taps, and K l is a set of different Doppler frequency shifts κ at the same time delay.

5. A signal detection method for an OTFS system according to claim 1, characterized in that, The said Step 4 includes: Calculate the channel complex gain of all transmission paths according to the standard V2V link path loss model. Then, based on the channel complex gain of different transmission paths, calculate the effective signal-to-interference-plus-noise ratio of each path, and determine the detection path and interference path according to the effective signal-to-interference-plus-noise ratio.

6. The signal detection method for an OTFS system according to claim 5, wherein, In the said Step 4, calculate the effective signal-to-interference-plus-noise ratio of the path in the following manner: Among which E S represents the signal transmission power, h i is the channel complex gain of the i-th transmission path, N0 represents the noise power, represents the set of channel matrix coefficients excluding h i The number of transmission paths is P, and P′ represents the number of detection paths.

7. A signal detection method for an OTFS system according to claim 4, characterized in that The said Step 5 includes: The two-dimensional data symbols in the receiver delay-time domain are obtained by partitioning in blocks according to the time dimension Similarly, two-dimensional data symbols in the time delay-time domain at the sending end are obtained. Obtain the input-output relationship in the delay-time domain of a single time block. wherein is the delay-time domain channel matrix of a single time block, is the noise vector; Substitute the input-output relationship into the least mean square error estimation formula to obtain Regard the above formula as a linear equation system Ax = b for solving, where Solve the system of equations by the conjugate gradient method to obtain the time-delay time-domain data symbols after the initial estimate And use it as the initial value for maximum ratio combining detection.

8. A signal detection method for an OTFS system according to claim 7, characterized in that, The estimated value of the transmitted signal is: In the formula, q = (m + nM) Among them is the received vector, the initial estimate value of is obtained from, where m = 0, 1, 2..., M - 1 and n = 0, 1, 2..., N - 1.

9. A signal detection method for an OTFS system according to claim 1, characterized in that The computational complexity of the multi-tap delay-time domain block decomposition least mean square error equalizer is O(KNM(2α - 1)), where K is the condition number of the covariance matrix of the covariance matrix, is the delay-time domain channel matrix of a single time block, α is the number of non-zero elements in each row of the covariance matrix, and α ≤ L, K and α are much smaller than NM.

Images