A joint lmmse and mrc orthogonal time frequency space modulation system symbol detection method

Abstract

This invention relates to a symbol detection method for an orthogonal time-frequency spatial modulation system combining LMMSE and MRC, belonging to the field of mobile communication. Based on the orthogonal time-frequency spatial OTFS modulation system, linear minimum mean square error (LMMSE) detection is first performed in blocks in the time domain. The detection results are then combined and converted to the time-delay-Doppler domain for orthogonal amplitude demodulation. The demodulated results are then converted to the time-delay-time domain and used as the initial values ​​for the maximum ratio combining (MRC) detection algorithm in the time-delay-time domain, thereby improving the bit error rate performance of the MRC detection method. Simulation experiments verify the effectiveness of this invention. Compared with other initialization detection methods, this method has a lower bit error rate at high communication rates and also exhibits superior performance in high-speed mobile scenarios.

Description

Technical Field

[0001] This invention belongs to the field of mobile communications and relates to communication symbol detection technology for orthogonal time-frequency-space (OTFS) modulation systems. Background Technology

[0002] In recent years, with the gradual expansion of 5G mobile communication technology deployment, traditional modulation schemes such as Orthogonal Frequency Division Multiplexing (OFDM) have suffered from severe inter-carrier interference due to high Doppler frequency shift in highly mobile user application scenarios such as high-speed railways, autonomous vehicles, and drones, thus degrading the overall system performance. To reduce undesirable Doppler frequency shift and improve overall system performance, a new modulation scheme—Orthogonal Time-Frequency Space (OTFS) modulation—has recently emerged. This scheme multiplexes information symbols in the delay-Doppler domain instead of in the time-frequency domain as in traditional modulation techniques, resulting in significant performance improvements in high-speed mobile scenarios. Each symbol in an OTFS data frame experiences the same channel, thus exhibiting better resistance to time-frequency fading. Furthermore, the OTFS system represents a time-frequency bidispersive channel in the delay-Doppler domain, which can more fully reflect the actual geometric sparsity characteristics of the wireless channel. Research on symbol detection methods for OTFS systems is currently a major research focus in the field of mobile communications.

[0003] Before implementing symbol detection methods, it is necessary to construct discrete-time delay-Doppler channel models, discrete-time delay-time domain channel models, and discrete-time domain channel models. Most current research constructs these three channel models based on standard wireless mobile multipath propagation scenarios. After constructing the channel models, symbol detection begins. Currently, symbol detection methods for OTFS fall into two main categories: linear and nonlinear. Linear detection methods primarily employ minimum mean square error (MMSE) detection. This method is simple to implement but has high complexity and poor bit error rate performance. Nonlinear detection commonly uses message passing (MP) methods, which offer better bit error rate performance but are complex to implement. To improve the performance of OTFS systems, some researchers have proposed an iterative maximum ratio combining (MRC) detection method, which receives different signals diversity, achieving a lower bit error rate. However, existing MRC detection methods suffer from decreased bit error rate performance when using high-order quadrature amplitude modulation (QAM) because they do not employ initialization detection or time-frequency single-tap initialization detection. Summary of the Invention

[0004] This invention provides a symbol detection method for orthogonal time-frequency control systems that combines LMMSE and MRC, aiming to reduce the bit error rate of OTFS systems under high communication rates, addressing the shortcomings of existing MRC detection methods.

[0005] The technical solution adopted by this invention includes the following steps:

[0006] Step 1: Based on the orthogonal time-frequency space-time (OTFS) modulation system, the communication symbols are subjected to orthogonal amplitude modulation (QAM) in the time-delay-Doppler domain with M rows and N columns, and some symbols are set to zero.

[0007] Step 2: Perform an N-point inverse fast Fourier transform (IFFT) on the rows of data in the time-delay-Doppler domain to transform them into the time-delay-time domain. Serialize the symbols in the time-delay-time domain by column and send them as discrete-time signals.

[0008] Step 3: Based on the standard wireless mobile multipath propagation scenario, construct a discrete delay-Doppler channel model and convert it to obtain discrete delay-time domain and discrete time domain channel models for subsequent symbol detection.

[0009] Step 4: After receiving symbols in the discrete time domain, the receiving end performs linear minimum mean square error detection on a block-by-block basis according to the discrete time domain channel and the received discrete time domain signal, and combines the block detection results into a time domain detection matrix of M rows and N columns.

[0010] Step 5: Convert the time-domain detection matrix into a time-delay-Doppler domain matrix and perform QAM demodulation in the time-delay-Doppler domain. Since performing maximum ratio combining MRC detection in the time-delay-time domain can reduce its detection complexity, the demodulation result in the time-delay-Doppler domain is converted to the time-delay-time domain as the initial value for maximum ratio combining detection.

[0011] Step 6: In the delay-time domain, the above results are used as initial values ​​to perform a maximum ratio combining detection with low complexity. The detection results are then processed to finally obtain the demodulated signal at the receiving end.

[0012] In step 1 of this invention, an orthogonal time-frequency spatial modulation frame is set. This frame consists of a two-dimensional plane with N blocks and M subcarriers, corresponding to the number of columns and rows of the time-delay-Doppler two-dimensional plane, respectively. Orthogonal amplitude modulation is performed on N×M information symbols, and the last l of the two-dimensional plane is... max The row is set to 0, where l max The maximum time delay propagation index (MTP) represents the propagation index of the channel. The frame format is set, and the duration of a single OTFS symbol is T. A frame has a total of N symbols, hence the frame duration T. f =NT, the system bandwidth B = MΔf, where Δf is the subcarrier spacing, using spacing T s =T / M, let T△f = 1.

[0013] In step 2 of this invention, an N-point inverse fast Fourier transform is performed on the rows of data in the time-delay-Doppler domain to transform them into the time-delay-time domain. The symbols in the time-delay-time domain are then serialized column-wise and transmitted as discrete-time signals. Let the two-dimensional transmitted symbol matrices in the time-delay-Doppler domain and the time-delay-time domain be respectively... The corresponding symbol vectors are respectively They can be converted to each other using a Fourier transform matrix, and the conversion method is as follows:

[0014]

[0015] Where F N This represents the normalized N-point discrete Fourier transform matrix. This indicates the conjugate transpose. After converting the transmitted symbol from the time-delay-Doppler domain to the time-delay-time domain, a parallel-to-serial conversion of the time-delay-time domain symbol yields the transmitted signal in the time domain. The transformation method is as follows:

[0016]

[0017] Among them, G tx For pulse shaping waveforms at the transmitting end, rectangular pulses are generally used, so it is equivalent to an identity matrix I of size M×M. M In this case, the transmitted discrete-time domain signal samples are associated with the delay-Doppler domain information symbols as follows:

[0018]

[0019] This gives us the time-domain transmission symbol, which is then transmitted from the transmitter.

[0020] In step 3 of this invention, a discrete-time delay-Doppler channel model is constructed based on a standard wireless mobile multipath propagation scenario. This model is then used to convert the discrete-time delay-time domain and discrete-time domain channel models. This step primarily constructs three channel models to prepare for subsequent symbol detection. These three channel models are the discrete-time domain channel model, the discrete-time delay-Doppler channel model, and the discrete-time delay-time domain channel model. Starting with the delay-Doppler channel model, consider a total of P paths, where the delay τ on the i-th path... i and Doppler frequency shift ν i They are respectively

[0021]

[0022] in and κ i These represent discrete time delay and discrete Doppler, respectively, τ max and ν maxRepresenting the maximum time delay and Doppler shift, the corresponding sparse representation of the channel is:

[0023]

[0024] Where δ(·) represents the Dirichlet function, h i The time delay is τ i Doppler frequency shift is ν i The channel complex gain corresponding to the time can be expressed as follows after discretization:

[0025]

[0026] in,

[0027]

[0028] and These are the corresponding discrete time delay set and discrete Doppler set, respectively;

[0029] Integrating the sparse representation h(τ,ν) of the channel along the Doppler axis yields the continuous time-delay channel impulse response as follows:

[0030]

[0031] After discretization and delay, we get:

[0032]

[0033] Further discretizing the time, we sample MN points from a frame, where 0 ≤ q ≤ NM-1, and the sampling time interval is T / M. Therefore, t = qT / M. Based on TΔf = 1, the above equation can be transformed into a discrete delay-time channel model:

[0034]

[0035] in l is an integer delay tap;

[0036] The relationships between time-domain and time-delay-time-domain transmitted signals and between time-domain and time-delay-time-domain received signals are as follows:

[0037]

[0038]

[0039] Let q = (m + nM), where m = 0, 1, ..., M-1, n = 0, 1, ..., N-1, then the received signal can be written as:

[0040]

[0041] Combining the above equation, we can obtain:

[0042]

[0043] in, Further order:

[0044]

[0045] The delay-time domain input / output expression for each block is then obtained as follows:

[0046]

[0047] in, The Hadamard product represents the matrix multiplication, which is the element-wise multiplication of the matrices.

[0048] From the input-output relationship in the discrete-time domain, the relationship between each input-output block in the discrete-time-Doppler domain can be obtained as follows:

[0049]

[0050] in, make:

[0051]

[0052] The input-output matrix form between each block in the discrete-time delay-Doppler domain is obtained as follows:

[0053]

[0054] Correspondingly, the input-output relationship of all blocks in a frame can be obtained, expressed as:

[0055] y = H·x + w

[0056] in w represents noise. For the discrete-time delay-Doppler channel model, and based on:

[0057]

[0058] in, Therefore, the input-output relationship of all blocks in a frame in the discrete time delay domain is obtained as follows:

[0059]

[0060] in, For discrete time delay - noise in the time domain, This is the discrete-time delay channel model;

[0061] Similarly, according to:

[0062]

[0063]

[0064]

[0065]

[0066] Where P is the permutation matrix, substituting it into the input-output relationship of all blocks in a frame in the discrete-time domain, we obtain the input-output relationship in the discrete-time domain of a frame as follows:

[0067]

[0068]

[0069]

[0070] Where T represents the transpose and G is the required discrete-time channel model.

[0071] In step 4 of this invention, linear minimum mean square error detection is performed in blocks based on the discrete-time channel and the received discrete-time domain signal. The detection result for the nth time-domain transmitted signal is as follows:

[0072]

[0073] Using the above method, only linear minimum mean square error detection needs to be applied to each block. For each block, the matrix dimension is M×M, and the time complexity of the corresponding linear minimum mean square error detection is O(M). 3 There are a total of N blocks, so the overall time complexity is O(NM). 3 This can greatly reduce the complexity of initial detection, and then the results of block-by-block detection can be processed. The columns are combined to form an M-row, N-column matrix S.

[0074] In step 5 of this invention, the time-domain detection results are processed in blocks. The matrix S formed is converted into a time-delay-Doppler domain matrix. The corresponding conversion relationship is:

[0075]

[0076] Detection signal in the time-delay-Doppler domain QAM demodulation is performed and used as the initial value for maximum ratio merge detection. Since detection in the delay-time domain can reduce the complexity of maximum ratio merge detection, it is also converted to the delay-time domain.

[0077] In step 6 of this invention, the time-delay-Doppler domain maximum ratio combining detection algorithm, for each received block y m+l Let (m∈0,1,...,M', l∈0,1,...,l) max M' is the number of non-zero blocks in the emitter block, l max (where x is the maximum delay propagation index of the channel), it is determined by the transmit block x m It is composed of different channel delays, and its specific expression is:

[0078]

[0079] in The different channel delays are caused by multipath propagation, and w is the corresponding noise. Given the channel and transmit block, for each receive block, the receive block with different delay components removed is calculated. Its expression is:

[0080]

[0081] In the above formula This is the result of the initial detection. Then, based on the maximum ratio, the detection is merged. For each transmitted detection block, all corresponding delay components are added together to obtain the detection result for each block:

[0082]

[0083] in, c m For each block detection result, perform maximum likelihood analysis on each symbol in the block detection result to obtain the detection value for each block:

[0084]

[0085] Among them, c m This is the output vector of the maximum ratio merging detection algorithm, n = 0, ..., N-1, a j It is a QAM modulated constellation diagram The symbols in the text are j = 1, ..., Q, where Q is the order of QAM modulation;

[0086] To further reduce the complexity of maximum ratio merging detection, it can be performed directly in the discrete time delay domain. First, the residual plus noise is defined as:

[0087]

[0088] Therefore, for the i-th iteration, the above-mentioned It can be rewritten as:

[0089]

[0090] For the first iteration, This is the result of block-based linear minimum mean square error detection. Compared to other initialization methods, this method has better detection performance and can reduce the number of iterations. This is based on the above formula and the channel relationship between the discrete delay-time domain and the discrete delay-Doppler domain derived in step 3. The maximum ratio of merging detection with reduced complexity can be obtained as follows:

[0091]

[0092] in,

[0093]

[0094]

[0095] Hadamard division represents the division of a matrix by corresponding elements. Defined as:

[0096]

[0097] when When the norm no longer decreases, that is Stop iterating, and then perform maximum likelihood detection for each symbol in the detection results of each block:

[0098]

[0099] Finally, the demodulated signal at the receiving end is obtained.

[0100] The advantage of this invention is that, based on the OTFS system architecture, orthogonal amplitude modulation is performed in the time-delay-Doppler domain. Some symbols are set to zero, and the cyclic characteristics of the channel matrix blocks are utilized to redefine the OTFS input-output relationship in a simple vector form. At the receiver, linear minimum mean square error (LMMSE) detection is performed block-by-block in the time domain instead of detecting the entire matrix, thus reducing the complexity of initial detection. The result is then transformed to the time-delay-Doppler domain for demodulation, and the demodulated result is converted to the time-delay domain as the initial value for maximum ratio combining (MRC) detection. This not only reduces the complexity of MRC detection but also improves the bit error rate performance of the MRC method under high-order orthogonal amplitude modulation. Experimental results show that, with the increase of the orthogonal amplitude modulation order, the method used in this invention has the lowest bit error rate compared to other methods. Furthermore, with the increase of user mobility speed, the bit error rate of the proposed method decreases, demonstrating the higher overall performance of the OTFS system in high-mobility application scenarios. Attached Figure Description

[0101] Figure 1 This is a flowchart of the orthogonal time-frequency control system symbol detection method combining LMMSE and MRC in this invention;

[0102] Figure 2(a) shows the matrix representation of the time delay-Doppler domain channel in the orthogonal time-frequency control system of the present invention;

[0103] Figure 2(b) shows the matrix representation of the time delay-time domain channel in the orthogonal time-frequency control system of the present invention;

[0104] Figure 2(c) shows the matrix representation of the time-domain channel in the orthogonal time-frequency control system of the present invention;

[0105] Figure 3 This is a schematic diagram of the time-delay-Doppler domain maximum ratio merging detection principle;

[0106] Figure 4 This is a comparison chart of the bit error rate of the proposed method and other OTFS symbol detection methods under the conditions of signal-to-noise ratio of 5dB to 25dB and 16-QAM modulation.

[0107] Figure 5(a) is a comparison of the bit error rate of the block-based LMMSE as MRC initialization detection method proposed in this invention and other methods using different initialization detection methods when the signal-to-noise ratio is 10dB to 30dB and the modulation order is 4-QAM.

[0108] Figure 5(b) is a comparison of the bit error rate of the block-based LMMSE as MRC initialization detection method proposed in this invention and other methods using different initialization detection methods when the signal-to-noise ratio is 10dB to 30dB and the modulation order is 8-QAM.

[0109] Figure 5(c) is a comparison of the bit error rate of the block-based LMMSE as MRC initialization detection method proposed in this invention and other methods using different initialization detection methods when the signal-to-noise ratio is 10dB to 30dB and the modulation order is 16-QAM.

[0110] Figure 6 This is a comparison chart of the bit error rate of the OTFS system under different user movement scenarios with a signal-to-noise ratio of 5dB to 25dB and a modulation order of 16-QAM. Detailed Implementation

[0111] like Figure 1 As shown, it includes the following steps:

[0112] Step 1: Based on the orthogonal time-frequency space-time (OTFS) modulation system, the communication symbols are subjected to orthogonal amplitude modulation (QAM) in the time-delay-Doppler domain with M rows and N columns, and some symbols are set to zero.

[0113] Step 2: Perform an N-point inverse fast Fourier transform (IFFT) on the rows of data in the time-delay-Doppler domain to transform them into the time-delay-time domain. Serialize the symbols in the time-delay-time domain by column and send them as discrete-time signals.

[0114] Step 3: Based on the standard wireless mobile multipath propagation scenario, construct a discrete delay-Doppler channel model and convert it to obtain discrete delay-time domain and discrete time domain channel models for subsequent symbol detection.

[0115] Step 4: After receiving symbols in the discrete time domain, the receiving end performs linear minimum mean square error detection on a block-by-block basis according to the discrete time domain channel and the received discrete time domain signal, and combines the block detection results into a time domain detection matrix of M rows and N columns.

[0116] Step 5: Convert the time-domain detection matrix into a time-delay-Doppler domain matrix and perform QAM demodulation in the time-delay-Doppler domain. Since performing maximum ratio combining MRC detection in the time-delay-time domain can reduce its detection complexity, the demodulation result in the time-delay-Doppler domain is converted to the time-delay-time domain as the initial value for maximum ratio combining detection.

[0117] Step 6: In the delay-time domain, the above results are used as initial values ​​to perform a maximum ratio combining detection with low complexity. The detection results are then processed to finally obtain the demodulated signal at the receiving end.

[0118] In step 1, an orthogonal time-frequency space-time (OTFS) modulation frame is set. This frame is composed of a two-dimensional plane, with N blocks and M subcarriers, corresponding to the number of columns and rows of the time-delay-Doppler two-dimensional plane, respectively. Orthogonal amplitude modulation is performed on N×M information symbols, and the last l of the two-dimensional plane is... max The row is set to 0, where l max This represents the maximum delay propagation index of the channel. Although setting it to zero will reduce the system's communication rate, it can avoid inter-block interference in the time domain and also help reduce the complexity of OTFS detection. Additionally, it sets the frame format; the duration of a single OTFS symbol transmitted is T, and a frame has a total of N symbols, hence the frame duration T. f =NT, the system bandwidth B = MΔf, where Δf is the subcarrier spacing and the sampling interval T s =T / M, let T△f = 1.

[0119] In step 2, an N-point inverse fast Fourier transform is performed on the rows of data in the time-delay-Doppler domain to transform them into the time-delay-time domain. The symbols in the time-delay-time domain are then serialized column-wise and transmitted as discrete-time signals. Let the two-dimensional transmitted symbol matrices in the time-delay-Doppler domain and the time-delay-time domain be respectively... The corresponding symbol vectors are respectively They can be converted to each other using a Fourier transform matrix, and the conversion method is as follows:

[0120]

[0121] Where F N This represents the normalized N-point discrete Fourier transform matrix. Representing the conjugate transpose, after converting the transmitted symbol from the time-delay-Doppler domain to the time-delay-time domain, a parallel-to-serial conversion of the time-delay-time symbol yields the transmitted signal in the time domain. The transformation method is as follows:

[0122]

[0123] Among them, G tx For pulse shaping waveforms at the transmitting end, rectangular pulses are generally used, so it is equivalent to an identity matrix I of size M×M. M In this case, the transmitted discrete-time domain signal samples are associated with the delay-Doppler domain information symbols as follows:

[0124]

[0125] This gives us the time-domain transmission symbol, which is then transmitted from the transmitter.

[0126] In step 3, a discrete-time delay-Doppler channel model is constructed based on the standard wireless mobile multipath propagation scenario. This model is then used to convert the discrete-time delay-time domain and discrete-time domain channel models. This step primarily constructs three channel models to prepare for subsequent symbol detection. These three channel models are the discrete-time domain channel model, the discrete-time delay-Doppler channel model, and the discrete-time delay-time domain channel model, as shown in Figures 2(a), 2(b), and 2(c). Starting with the delay-Doppler channel model, consider a total of P paths, where the delay τ on the i-th path... i and Doppler frequency shift ν i They are respectively:

[0127]

[0128] in and κ i These represent discrete time delay and discrete Doppler, respectively, τ max and ν max Representing the maximum time delay and Doppler shift, the corresponding sparse representation of the channel is:

[0129]

[0130] Where δ(·) represents the Dirichlet function, hi The time delay is τ i Doppler frequency shift is ν i The channel complex gain corresponding to the time can be expressed as follows after discretization:

[0131]

[0132] in,

[0133]

[0134] and These are the corresponding discrete time delay set and discrete Doppler set, respectively;

[0135] Integrating the sparse representation h(τ,ν) of the channel along the Doppler axis yields the continuous time-delay channel impulse response as follows:

[0136]

[0137] After discretization and delay, we get:

[0138]

[0139] Further discretizing the time, we sample MN points from a frame, where 0 ≤ q ≤ NM-1, with a sampling time interval of T / M. Therefore, t = qT / M. Based on TΔf = 1, the above equation can be transformed into a discrete-time delay channel model:

[0140]

[0141] in l is an integer delay tap;

[0142] The relationships between time-domain and time-delay-time-domain transmitted signals and between time-domain and time-delay-time-domain received signals are as follows:

[0143]

[0144]

[0145] Let q = (m + nM), where m = 0, 1, ..., M-1, n = 0, 1, ..., N-1, then the discrete-time received signal can be written as:

[0146]

[0147] Combining the above equations, we can obtain the input-output relationship in the discrete-time domain and the time domain as follows:

[0148]

[0149] in Further order:

[0150]

[0151] The discrete-time delay-time domain input-output expression for each block is then obtained as follows:

[0152]

[0153] in, The Hadamard product represents the matrix multiplication, which is the element-wise multiplication of the matrices.

[0154] From the input-output relationship of each block in the discrete-time domain, the relationship of each input-output block in the discrete-time-Doppler domain can be obtained as follows:

[0155]

[0156] in, make:

[0157]

[0158] The input-output matrix form between each block in the discrete-time delay-Doppler domain is obtained as follows:

[0159]

[0160] Correspondingly, the input-output relationship of all blocks in a frame can be obtained, as shown in Figure 2(a), which is represented as:

[0161] y = H·x + w

[0162] in w represents noise. This is the required discrete-time delay-Doppler channel model, and then based on:

[0163]

[0164] in, Therefore, the input-output relationship of all blocks in a frame in the discrete time delay domain is obtained as follows:

[0165]

[0166] in, For discrete time delay - noise in the time domain, This is the time-delay-time domain channel model;

[0167] Similarly, according to:

[0168]

[0169]

[0170]

[0171]

[0172] Where P is the permutation matrix, substituting it into the input-output relationship of all blocks in a frame in the discrete-time domain, we obtain the input-output relationship in the discrete-time domain of a frame as follows:

[0173]

[0174]

[0175]

[0176] Where T represents the transpose and G is the required discrete-time channel model.

[0177] In step 4, after receiving symbols in the time domain, the receiving end performs linear minimum mean square error detection on a block-by-block basis according to the discrete-time channel and the received discrete-time signals. The detection result for the nth time-domain transmitted signal is as follows:

[0178]

[0179] If we directly perform linear minimum mean square error detection on the received time signal, we need to invert an MN×MN matrix, and its time complexity is O(M). 3 N 3 Therefore, by using the above method, we only need to apply linear minimum mean square error detection to each block. For each block, the matrix dimension is M×M, and the time complexity of the corresponding linear minimum mean square error detection is O(M). 3 There are a total of N blocks, so the overall time complexity is O(NM). 3 This can greatly reduce the complexity of initial detection, and then the results of block-by-block detection can be processed. The columns are combined to form an M-row, N-column matrix S;

[0180] Step 5 involves dividing the time-domain detection results into blocks. The matrix S formed is converted into a time-delay-Doppler domain matrix. The corresponding conversion relationship is:

[0181]

[0182] Detection signal in the time-delay-Doppler domain QAM demodulation is performed and used as the initial value for maximum ratio merge detection. Since detection in the delay-time domain can reduce the complexity of maximum ratio merge detection, it is also converted to the delay-time domain.

[0183] In step 6, in the time-delay domain, the above results are used as initial values ​​for maximum ratio merging detection with reduced complexity. First, the maximum ratio merging detection algorithm in the time-delay-Doppler domain is as follows: Figure 3 As shown, for each receiving block y m+l Let (m∈0,1,...,M', l∈0,1,...,l) max M' is the number of non-zero blocks in the emitter block, l max (where x is the maximum delay propagation index of the channel), it is determined by the transmit block x m It is composed of different channel delays, and its specific expression is:

[0184]

[0185] in The different channel delays are caused by multipath propagation, where w is the corresponding noise. Given the known channel and transmit block, for each receive block, the receive block with different delay components removed is calculated. Its expression is:

[0186]

[0187] For example The x2 transmitting block is received by the y3 receiving block after passing through a path with a channel delay exponent of 1. In the above formula... This is the result of the initial detection. Then, based on the maximum ratio, the detection is merged. For each transmitted detection block, all corresponding delay components are added together to obtain the detection result for each block:

[0188]

[0189] in, c m For each block detection result, perform maximum likelihood analysis on each symbol in the block detection result to obtain the detection value for each block:

[0190]

[0191] Among them, c m This is the output vector of the maximum ratio merging detection method, n = 0, ..., N-1, a j It is a QAM modulated constellation diagram The symbols in the text are j = 1, ..., Q, where Q is the order of QAM modulation;

[0192] To further reduce the complexity of maximum ratio merging detection, it can be performed directly in the discrete time delay domain. First, the residual plus noise is defined as:

[0193]

[0194] Therefore, for the i-th iteration, the above-mentioned It can be rewritten as:

[0195]

[0196] For the first iteration, This is the result of block-based linear minimum mean square error detection. Compared to other initialization methods, this method has better detection performance and can reduce the number of iterations. This is based on the above formula and the channel relationship between the discrete delay-time domain and the discrete delay-Doppler domain derived in step 3. The maximum ratio of merging detection with reduced complexity can be obtained as follows:

[0197]

[0198] in,

[0199]

[0200]

[0201] Hadamard division represents the division of a matrix by corresponding elements. Defined as:

[0202]

[0203] when When the norm no longer decreases, that is Stop iterating, and then perform maximum likelihood detection for each symbol in the detection results of each block:

[0204]

[0205] Finally, the demodulated signal at the receiving end is obtained.

[0206] The beneficial effects of the present invention will be further explained below with reference to simulation experiments and results. The simulation tool used for all simulation experiments is Matlab.

[0207] Experiment 1: The bit error rate of the method of the present invention is compared with that of the existing MP method, time-frequency single-tap method, and time-domain block LMMSE method under the conditions of 5dB to 25dB signal-to-noise ratio.

[0208] Simulation parameter settings: Set frame size M=N=8, carrier frequency to 4×10⁻⁶. 9 The frequency is Hz, the subcarrier spacing is 15000Hz, the QAM modulation order is 32-QAM, the signal-to-noise ratio is 5dB to 25dB, the maximum channel delay tap is 3, and the number of frames is 1000.

[0209] Depend on Figure 4 As can be seen, under 16-QAM modulation, compared with other detection methods of OTFS, the present invention first uses block-based linear minimum mean square error detection as the initial detection, and then combines it with the maximum ratio merging detection in the time-delay domain to achieve the lowest bit error rate, thus verifying the superiority of the method of the present invention.

[0210] Experiment 2: The bit error rate of the method of the present invention was compared with that of the maximum ratio merging detection method using time-frequency single-tap detection as initialization and without initialization under different orthogonal amplitude modulation.

[0211] Simulation parameter settings: Set frame size M=N=8, carrier frequency to 4×10⁻⁶. 9 The frequency is Hz, the subcarrier spacing is 15000Hz, the QAM modulation order is 4-QAM, 8-QAM, 16-QAM, the signal-to-noise ratio is 10dB to 30dB, the maximum channel delay tap is 3, and the number of frames is 2000.

[0212] As can be seen from Figures 5(a), 5(b), and 5(c), when the signal-to-noise ratio is 10dB to 30dB, the method used in this invention has the lowest bit error rate as the quadrature amplitude modulation order increases, verifying the superiority of the method proposed in this invention compared with other initialization methods.

[0213] Experiment 3: To verify the performance of the method proposed in this invention in high-mobility wireless channels, the bit error rate was compared at different user movement speeds.

[0214] Simulation parameter settings: Set frame size M=N=8, carrier frequency to 4×10⁻⁶. 9 The frequency is Hz, the subcarrier spacing is 15000Hz, the QAM modulation order is 16-QAM, the signal-to-noise ratio is 5dB to 25dB, the standard EVA channel is used, and the number of frames is 2000.

[0215] Depend on Figure 6 It can be seen that when the speed is below 1000 km / h, the bit error rate of the method of this invention gradually decreases as the user's moving speed increases. This is because when the Doppler diffusion is greater than the Doppler resolution, different paths located within the same delay tap are separated into different Dopplers. When the speed is greater than 1000 km / h, all paths are divided into different delay-Dopplers, and at this point, the bit error rate no longer decreases with increasing speed.

[0216] This invention proposes a novel symbol detection method for orthogonal time-frequency control systems. First, a block-based linear minimum mean square error (MMS) method is employed in the time domain to transform the detection results into the time-delay-Doppler domain for demodulation. The demodulated symbols are then transformed back into the time-delay-time domain as initial values ​​for the maximum ratio combining (MRC) detection method, thereby improving the bit error rate performance of MRC detection. Simulation experiments verify that the proposed method exhibits better bit error rate performance than other detection algorithms in OTFS systems, and also demonstrates superior performance in high-speed mobile scenarios.

Claims

1. A symbol detection method for an orthogonal time-frequency control system combining LMMSE and MRC, characterized in that, Includes the following steps: Step 1: Based on the orthogonal time-frequency space-time (OTFS) modulation system, the communication symbols are subjected to orthogonal amplitude modulation (QAM) in the time-delay-Doppler domain with M rows and N columns, and some symbols are set to zero. Step 2: Perform an N-point inverse fast Fourier transform (IFFT) on the rows of data in the time-delay-Doppler domain to transform them into the time-delay-time domain. Serialize the symbols in the time-delay-time domain by column and send them as discrete-time signals. Step 3: Based on the standard wireless mobile multipath propagation scenario, construct a discrete delay-Doppler channel model and convert it to obtain discrete delay-time domain and discrete time domain channel models for subsequent symbol detection. Step 4: After receiving symbols in the discrete time domain, the receiving end performs linear minimum mean square error detection on a block-by-block basis according to the discrete time domain channel and the received discrete time domain signal, and combines the block detection results into a time domain detection matrix of M rows and N columns. Step 5: Convert the time-domain detection matrix into a time-delay-Doppler domain matrix, perform QAM demodulation in the time-delay-Doppler domain, and then convert the demodulation result in the time-delay-Doppler domain to the time-delay-time domain as the initial value for maximum ratio merging detection; Step 6: In the time-delay domain, the demodulation result obtained in step 5 is used as the initial value to perform a maximum ratio combining detection with low complexity. The detection result is then processed to finally obtain the demodulated signal at the receiving end.

2. The symbol detection method for an orthogonal time-frequency control system combining LMMSE and MRC according to claim 1, characterized in that: In step 1, an orthogonal time-frequency spatial modulation frame is set. This frame consists of a two-dimensional plane, with N blocks and M subcarriers, corresponding to the number of columns and rows of the delay-Doppler two-dimensional plane, respectively. Orthogonal amplitude modulation is applied to each information symbol, and the end of this two-dimensional plane is... The row is set to 0, where This represents the maximum time delay propagation index of the channel, sets the frame format, and specifies the duration of a single OTFS symbol during transmission. A frame contains a total of N symbols, therefore the duration of a frame is... System bandwidth ,in For subcarrier spacing, use interval ,make .

3. The symbol detection method for an orthogonal time-frequency control system combining LMMSE and MRC according to claim 1, characterized in that: In step 2, an N-point inverse fast Fourier transform is performed on the rows of data in the time-delay-Doppler domain to transform them into the time-delay-time domain. The symbols in the time-delay-time domain are then serialized column-wise and transmitted as discrete-time signals. Let the two-dimensional transmitted symbol matrices in the time-delay-Doppler domain and the time-delay-time domain be respectively... , The corresponding symbol vectors are respectively , They can be converted to each other using Fourier transform matrices, and the conversion method is as follows: ； in This represents the normalized N-point discrete Fourier transform matrix. This indicates the conjugate transpose. After converting the transmitted symbol from the time-delay-Doppler domain to the time-delay-time domain, a parallel-to-serial conversion is performed on the time-delay-time domain symbol to obtain the transmitted signal in the time domain. The transformation method is as follows: ； in, To shape the pulse waveform at the transmitting end, a rectangular pulse is used, so it is equivalent to a large / small pulse. identity matrix The transmitted discrete-time domain signal samples are associated with the delay-Doppler domain information symbols as follows: ； This gives us the time-domain transmission symbol, which is then transmitted from the transmitter.

4. The symbol detection method for an orthogonal time-frequency control system combining LMMSE and MRC according to claim 1, characterized in that: In step 3, a discrete-time delay-Doppler channel model is constructed based on the standard wireless mobile multipath propagation scenario. This model is then transformed into discrete-time delay-time domain and discrete-time domain channel models. In this step, three channel models are constructed to prepare for subsequent symbol detection. These three channel models are the discrete-time domain channel model, the discrete-time delay-Doppler channel model, and the discrete-time delay-time domain channel model. Starting with the delay-Doppler channel model, consider a total of P paths, where the... Delay on the path and Doppler shift They are respectively: ， ； in and These are discrete time delay and discrete Doppler, respectively. and Representing the maximum time delay and Doppler shift, the corresponding sparse representation of the channel is: ； in Represents the Dirichlet function, Indicates the delay as Doppler frequency shift is The corresponding channel complex gain, after discretization, is expressed as: ； in, ； and These are the corresponding discrete time delay set and discrete Doppler set, respectively; Sparse representation of the channel Integrating along the Doppler axis, the continuous time-delay channel impulse response is obtained as follows: ； After discretization and delay, we get: ； Further discretize the time by sampling MN points from the time of a frame, where The sampling time interval is Therefore And then according to The above equation is transformed into a discrete-time delay-time channel model: ； in , For integer delay taps; The relationships between time-domain and time-delay-time-domain transmitted signals and between time-domain and time-delay-time-domain received signals are as follows: ； make ,in , The received signal is then written as: ； Combining the above equation, we get: ； in, , and then: ； The delay-time domain input / output expression for each block is then obtained as follows: ； in, The Hadamard product represents the matrix multiplication, which is the element-wise multiplication of the matrices. From the input-output relationship in the discrete-time-delay domain, the relationship between each input-output block in the discrete-time-delay-Doppler domain is obtained as follows: ； in, ,make: ； The input-output matrix form between each block in the discrete-time delay-Doppler domain is obtained as follows: ； Correspondingly, the input-output relationship of all blocks in a frame is obtained, expressed as: ； in , For noise, For the discrete-time delay-Doppler channel model, and based on: ； in, Therefore, the input-output relationship of all blocks in a frame in the discrete time delay domain is obtained as follows: ； in, , , For discrete time delay - noise in the time domain, This is the discrete-time delay channel model; Similarly, according to: ； ； ； in It is a permutation matrix. Substituting it into the input-output relationship of all blocks in a frame in the discrete-time domain, we obtain the input-output relationship in the discrete-time domain of a frame as follows: ； ； ； in, Indicates transpose. This is the required discrete-time channel model.

5. The method for symbol detection of an orthogonal time-frequency control system combining LMMSE and MRC according to claim 1, characterized in that, In step 4, linear minimum mean square error detection is performed in blocks based on the discrete-time channel and the received discrete-time domain signal. The detection result for the nth time-domain transmitted signal is as follows: ； Linear minimum mean square error detection is applied to each block. For each block, the dimension of its matrix is... The corresponding time complexity of linear minimum mean square error detection is There are a total of N blocks, so the total time complexity is This reduces the complexity of initial detection, and then the results of block detection are processed. A matrix formed by combining columns into M rows and N columns. .

6. The symbol detection method for an orthogonal time-frequency control system combining LMMSE and MRC according to claim 1, characterized in that, In step 5, the time-domain block detection results are processed. The matrix formed Convert to time-delay-Doppler domain matrix The corresponding conversion relationship is: ； Detection signal in the time-delay-Doppler domain QAM demodulation is performed, and the result is used as the initial value for maximum ratio combining detection. This value is also converted to the delay-time domain.

7. The symbol detection method for an orthogonal time-frequency control system combining LMMSE and MRC according to claim 1, characterized in that: In step 6, the delay-time domain maximum ratio combining detection algorithm, for each received block... In other words, , , The number of non-zero blocks in the emit block. The maximum delay propagation index of the channel is given by the transmit block. It is composed of different channel delays, and its specific expression is: ； in The different channel delays are caused by multipath propagation. For the corresponding noise, given the channel and transmit block, for each receive block, the receive block with different delay components removed is calculated. Its expression is: ； In the above formula This is the result of the initial detection. Then, based on the maximum ratio, the detection is merged. For each transmitted detection block, all corresponding delay components are added together to obtain the detection result for each block: ； in, , , For each block detection result, perform maximum likelihood analysis on each symbol in the block detection result to obtain the detection value for each block: ； in, It is the output vector of the maximum ratio merging detection algorithm. , It is a QAM modulated constellation diagram The symbols in , It is the order of QAM modulation; In the maximum ratio merging detection in the discrete time-delay domain, the residual plus noise is first defined as: ； Therefore, for the first The next iteration, as mentioned above Rewritten as: ； For the first iteration, This is the result of block-based linear minimum mean square error detection. Compared to other initialization methods, this method has better detection performance and reduces the number of iterations. This is based on the above formula and the channel relationship between the discrete delay-time domain and the discrete delay-Doppler domain derived in step 3. The maximum ratio of reduced complexity obtained by merging detection is: ； in, ； ； Hadamard division represents the division of a matrix by corresponding elements. Defined as: ； when When the norm no longer decreases, that is Stop the iteration, and then perform maximum likelihood detection on each symbol in the detection results of each block: ； Finally, the demodulated signal at the receiving end is obtained.

Images