Upstream multi-user random modulation and oamp detection method

Abstract

The application discloses an uplink multi-user random modulation and OAMP detection method, which comprises the following steps: independently modulating and transmitting signals of each user, and receiving the sum of signals of each user; performing linear estimation according to prior distribution of the transmitted signals of each user, state information of a channel matrix, estimation error of the channel matrix and channel noise power; performing nonlinear estimation according to the prior distribution of the transmitted signals of each user and the estimation and variance obtained by linear estimation, so as to obtain the nonlinear estimation and variance in the present iteration; continuously iterating the linear estimation and the nonlinear estimation, and performing hard decision according to the linear estimation and variance obtained by the last iteration to obtain a final decoding result. The application independently modulates all users, considers the channel estimation error, does not require that the entire equivalent channel matrix has right unitary invariant characteristics, reduces performance loss, and obtains detection performance which is obviously superior to that of the prior OAMP.

Description

Technical Field

[0001] This invention relates to the field of signal detection, and more particularly to an uplink multi-user random modulation and OAMP detection method. Background Technology

[0002] Orthogonal Approximate Message Passing (OAMP) is a commonly used algorithm for signal detection at the receiver in communication scenarios. As a variant of the Approximate Message Passing (AMP) algorithm, OAMP overcomes the limitation of AMP, which is only applicable to independent and identically distributed (IID) Gaussian matrices, and extends the range of matrices applicable to right unitary invariant matrices.

[0003] However, even though the OAMP algorithm has expanded its applicability compared to the original AMP algorithm, it still cannot be well applied to real-world communication scenarios because the channel matrix in real-world communication scenarios often lacks the right-unitary invariant property. Therefore, to enable the OAMP algorithm to be applied to signal detection in real-world communication scenarios, random modulation techniques can be used to precode the transmitted signal. After random modulation, the system's equivalent channel matrix will possess the right-unitary invariant property, thus allowing the OAMP algorithm to be used for signal detection.

[0004] However, the OAMP algorithm described above is generally only used for single-user signal detection. In multi-user communication scenarios, users are relatively independent, and it is impossible to perform joint random modulation on the transmitted signals of all users. Each user needs to perform independent random modulation. Therefore, when the Channel State Information (CSI) among multiple users in the system is asymmetric, the system does not possess right-unitary invariance with respect to the entire equivalent channel matrix of all users, resulting in a significant performance loss for the OAMP algorithm. Furthermore, while the OAMP algorithm previously assumed accurate channel CSI for signal detection, in reality, the estimated channel CSI will always deviate from the true channel CSI. When the CSI deviation is significant, the OAMP algorithm will also experience a significant performance loss.

[0005] Currently, common multi-user signal detection methods often require using Orthogonal Frequency Division Multiple Access (OFDMA) technology at the transmitter to modulate the signals of each user, and then performing minimum mean square error (MMSE) detection at the receiver. However, this detection method is only applicable to static multipath channels. If the communication channel is a time-varying multipath channel, OFDMA technology cannot diagonalize the equivalent channel matrix, thus making MMSE detection impossible.

[0006] In uplink multi-user scenarios, it is impossible to perform joint random modulation for all users. Each user needs to perform independent random modulation. When the CSI between users is asymmetric, the equivalent channel matrix does not have the right unitary invariant property, and the OAMP algorithm cannot be applied.

[0007] When the estimated channel CSI deviates significantly from the actual channel CSI, the OAMP algorithm will suffer a significant performance loss.

[0008] Multi-user MMSE detection based on OFDMA technology will fail in time-varying multipath channels. Summary of the Invention

[0009] The purpose of this invention is to address the shortcomings of existing technologies by proposing an uplink multi-user random modulation and OAMP detection method.

[0010] The objective of this invention is achieved through the following technical solution: an uplink multi-user random modulation and OAMP detection method, comprising:

[0011] S1. Transmit the signal after performing independent random modulation on each user at the transmitting end;

[0012] S2. Receive the sum of signals from each user at the receiving end;

[0013] S3. Based on the received signal and the state information of the channel matrix, the estimation error of the channel matrix and the channel noise power, perform linear estimation;

[0014] The orthogonal parameters are calculated based on the channel state information, the estimation error of the channel matrix, and the channel noise power.

[0015] The linear estimator for the current iteration is calculated based on the orthogonal parameters, channel state information, estimation error of the channel matrix, channel noise power, and nonlinear estimator from the previous iteration.

[0016] The linear estimation variance in this iteration is obtained based on the orthogonal parameters, channel state information, estimation error of the channel matrix, channel noise power, and nonlinear estimation variance in the previous iteration.

[0017] S4. Based on the prior distribution of each user's transmitted signal, the estimator and variance obtained from the linear estimation are used for nonlinear estimation to obtain the nonlinear estimator and variance in this iteration; the linear and nonlinear estimations are continuously iterated until convergence or the maximum number of iterations is reached to obtain the linear estimator and variance of the last iteration.

[0018] S5. Make a hard decision based on the obtained linear estimator and variance to obtain the final decoding result.

[0019] On the other hand, this invention also provides an uplink multi-user random modulation and OAMP detection device, including a memory and one or more processors, wherein the memory stores executable code, characterized in that when the processor executes the executable code, it implements the aforementioned uplink multi-user random modulation and OAMP detection method.

[0020] On the other hand, this specification also provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the aforementioned uplink multi-user random modulation and OAMP detection method.

[0021] The advantages of this invention are as follows: Previous OAMP algorithms required joint random modulation for all users, which is impractical in real-world multi-user communication scenarios. This invention, however, requires each user to perform independent random modulation in multi-user communication scenarios, making it a more practical and feasible solution.

[0022] When users perform independent random modulation, if there is a significant asymmetry in the CSI between users, the entire equivalent channel matrix of the system no longer has the right unitary invariant property, and the ordinary OAMP detection technology will suffer a large performance loss. However, the MU-OAMP detection technology proposed in this invention does not require the entire equivalent channel matrix to have the right unitary invariant property, and its detection performance is significantly better than that of ordinary OAMP.

[0023] Previous OAMP algorithms assumed that the channel CSI was accurate when applied, but the detection performance would degrade significantly when the estimated channel CSI deviated significantly from the actual channel CSI. The MU-OAMP detection technology proposed in this invention takes channel estimation error into account and achieves significantly better detection performance than OAMP by modifying and improving the LMMSE filter.

[0024] Currently, MMSE detection based on OFDMA technology, which is commonly used for multi-user signal detection in the field of communication, can only be implemented under static channels. However, the MU-OAMP detection technology proposed in this invention can not only achieve better detection performance under dynamic channels, but also significantly outperforms the former even under static channels. Attached Figure Description

[0025] Figure 1 A schematic diagram of the multi-user random modulation process;

[0026] Figure 2 A schematic diagram of the multi-user OAMP (MU-OAMP) detection technology process;

[0027] Figure 3 A comparison of bit error rates between multi-user OAMP and ordinary OAMP systems with the same transmit length, assuming accurate channel estimation.

[0028] Figure 4 A comparison of bit error rates between multi-user OAMP and ordinary OAMP systems with different transmit lengths, assuming accurate channel estimation.

[0029] Figure 5 The figure shows a comparison of the system bit error rate when using an uncorrected LMMSE filter and a corrected LMMSE filter for signal recovery, where the channel estimation error matrix is ​​an IID Gaussian matrix.

[0030] Figure 6 The figure shows a comparison of the system bit error rate when using a matched filter and a modified LMMSE filter for signal recovery, where the channel estimation error matrix is ​​an IID Gaussian matrix.

[0031] Figure 7 The figure shows a comparison of the system bit error rate when the user transmitted signal length is different and the signal is recovered using an uncorrected LMMSE filter and a corrected LMMSE filter, both with the channel estimation error matrix being an IID Gaussian matrix.

[0032] Figure 8 The figure shows a comparison of the system bit error rate when the user transmitted signal length is different under the condition that the channel estimation error matrix is ​​IID Gaussian matrix, and the matching filter and the modified LMMSE filter are used for signal recovery respectively.

[0033] Figure 9 To ensure accurate channel estimation, a comparison of bit error rates between OFDMA technology and a multi-user OAMP system under stationary user conditions is presented.

[0034] Figure 10 This is a schematic diagram of the uplink multi-user random modulation and OAMP detection device provided in an embodiment of the present invention. Detailed Implementation

[0035] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0036] Example 1

[0037] like Figure 1 and Figure 2 As shown, the present invention provides an uplink multi-user random modulation and OAMP detection method. Before detection, each user needs to be independently random-modulated at the transmitting end before transmitting the signal, and the sum of the signals from each user is received at the receiving end. Here, we consider an uplink communication system with U users and random modulation:

[0038]

[0039] in It is the sum of the signals received by the base station from U users after passing through their respective channels and being affected by noise interference; It is the Gaussian channel noise vector of IID, with It is the transmitted signal vector from the u-th user. This represents the Multi-Input Multi-Output (MIMO) channel matrix corresponding to the u-th user, where dimension L is the number of subcarriers. Signal x u It is a randomly modulated signal, represented as

[0040] x u =Ξ u s u ,

[0041] in It is of dimension L×N u A random unitary matrix, It is a dimension of N u A × 1 unmodulated input signal vector. Input signal vectors s from U users. u and the random unitary matrix Ξ u Each of these must be independent of the others and generated independently. It is important to note that for the u-th user, when the equivalent channel matrix A... u =H u Ξ u Only when it belongs to the universal class of matrices can x be... u =Ξ u s u Called the pair of s u Random modulation. The universal class matrix must satisfy the following condition:

[0042] 1)A u It is spectrally convergent and possesses a bounded spectral norm, satisfying...

[0043] 2) For any fixed constant And constants ∈>0, we have

[0044]

[0045] Theoretically, only unitary matrices randomly selected from the entire feasible region are strictly optimal; however, experimental results show that most randomly generated unitary matrices are good modulation matrices, so easily implemented random unitary matrices can be chosen, such as:

[0046] Matrix Ξ u You can choose the QR decomposition matrix of either the Haar matrix or the IID Gaussian matrix;

[0047] Matrix Ξ u It can be achieved through random permutation matrix Π u With unitary matrix T u Constructed by multiplication, i.e., Ξ u =Π u T u T u You can choose the Discrete Fourier Transform (DFT) matrix, the Hadamard-Walsh Transform matrix, the Discrete Sine Transform matrix, and the Normalized IID Gaussian Transform matrix.

[0048] Matrix Ξ u Alternatively, it can be constructed using a multi-level row permutation unitary matrix, denoted as Ξ u =Π u,1 T u,1 Π u,2 T u,2 …Π u, Q T u,Q Each pair of {Π u,q ,T u,q All of them are randomly selected.

[0049] The MU-OAMP algorithm performs linear and nonlinear estimations based on the channel information of the received signals transmitted by each user; it iterates the linear and nonlinear estimations until convergence or the maximum number of iterations is reached, and obtains the linear estimate and variance of the last iteration.

[0050] For ease of explanation, the descriptions here refer only to the u-th user; the signal estimation methods for other users are exactly the same.

[0051] The MU-OAMP algorithm proposed in this application is suitable for signal detection when the channel CSI is not perfect. Let the channel matrix obtained by channel estimation be... The estimated equivalent channel matrix is ​​then... If the estimated channel deviates from the actual channel, and the channel estimation error matrix is ​​denoted as... The equivalent channel estimation error matrix is by Using the initial estimated value for the u-th user, the multi-user OAMP algorithm proposed in this application iterates as follows:

[0052]

[0053] Hard decision is made based on the obtained linear estimator and variance to obtain the final decoding result.

[0054] For normalized QPSK signals remember So The judgment result can be obtained from the following formula.

[0055]

[0056] in This represents the i-th element of the decoding decision result.

[0057]

[0058] For the linear estimator in the last iteration The i-th element, T is the maximum number of iterations set in advance or the number of iterations when the iteration converges.

[0059] Because in d k,q If the value is too large or too small, MATLAB will perform calculations... Numerical overflow issues may occur, so we assign d in the actual program. k,q The upper and lower bounds are 100 and -100 respectively, denoted as We have

[0060]

[0061] Example 2:

[0062] This embodiment is for situations where there is an error between the estimated channel CSI and the actual channel CSI, and the channel estimation error matrix is ​​an IID Gaussian matrix. The implementation is as follows: the communication channel is a time-varying multipath channel, and the CSI estimated by each user is different and may have significant differences; the CSI estimated by each user may have different degrees of error with their respective actual CSI, and the channel estimation error matrix is ​​an IID Gaussian matrix; the transmitted signal length of each user may be different.

[0063] Assume the base station receives normalized QPSK signals transmitted by two users (U=2), with signal lengths N1 and N2 transmitted by user 1 and user 2 respectively; let the signal lengths transmitted by user 1 and user 2 be N1 and N2 respectively. and Then there is

[0064] s1~ i.i.d. S1,

[0065] s2~ i.i.d. S2, where

[0066]

[0067] u = 1, 2,

[0068] I u Q represents the component that is in phase with the reference carrier. u The component orthogonal to the reference carrier is shown. It can be generated using MATLAB's built-in uniform distribution function, randi. u This is a constellation diagram of QPSK signals, where Unif(·) represents a uniform distribution;

[0069] For the sake of brevity, we assume that both User 1 and User 2 use the TDL-A channel in the 5G 3GPP standard (but in reality, the channel categories of the two users can be completely different).

[0070] Set the number of subcarriers to L, the subcarrier spacing to Δf, and the carrier frequency to f. c Therefore, the sampling frequency f can be calculated. s =LΔf, sampling interval T s =1 / f s Set the total number of sampling symbols to s. pa The oversampling rate is s ps ;

[0071] Assuming the relative velocities between user 1 and user 2 and the base station are v1 and v2 respectively, the maximum Doppler frequency shifts of user 1 and user 2 can be calculated as ν. max,1 =f c v1 / c and ν max,2 =f c v² / c, c = 3e 8 m / s represents the speed of light.

[0072] Assume that the power gains of user 1 and user 2 due to slow fading of the channel are p1 and p2, respectively;

[0073] The latency spreads for users 1 and 2 are d1 and d2, respectively.

[0074] Generate the channel matrix for the u-th user (u = 1, 2) based on the following channel model.

[0075]

[0076] in

[0077]

[0078] l, p are indices, τ k,u Let v be the latency of the k-th path for the u-th user. k,u For the Doppler frequency shift of the u-th user and the k-th path,

[0079] Detailed explanation:

[0080] The parameters involved in the above channel model are:

[0081]

[0082] τ k,u =τ k,0 ·d u ,

[0083]

[0084]

[0085] The channel impulse response of the u-th user is:

[0086]

[0087] The transmitted signal passes through a transmission filter h rc (t), the response of the entire equivalent channel is:

[0088]

[0089] Where K is the number of channel paths, v k , τ k These represent the Doppler frequency shift and time delay of the k-th path, respectively; It is the time-varying complex amplitude gain of the k-th path, φ k,u (t) is a random phase; h rc (t) is a raised cosine filter (RCF):

[0090]

[0091] Where β∈[0,1] is the roll-off factor, T s It is the symbol interval / sampling interval. Assume the signal is oversampled after passing through this filter, and the total number of sampled symbols is s. pa The oversampling rate is sps Then the filter order is s pa ·s ps +1,

[0092] The actual amplitude gain of the u-th user |h k,u (t)| and normalized amplitude gain|h k,0 (t)| is directly proportional. To distinguish the channels of different users, we assign different power gains p to the signals received by different users at the receiver. u To simulate the differences in slow fading caused by varying distances, i.e.

[0093]

[0094] u = 1, ..., U.

[0095] The actual delay τ of the u-th user k,u Also related to the normalized time delay τ k,0 They are directly proportional. Assume the channel delay spread for the u-th user is d. u The actual delay can then be calculated using the following conversion formula.

[0096] τ k,u =τ k,0 ·d u ,

[0097] u = 1, ..., U.

[0098] Normalized delay τ of TDL channel k,0 (k=1,…,K), normalized power gain |h k,0 (t)| 2 These parameters can all be found in the 5G 3GPP standard, and the 5G toolbox provided by MATLAB can also directly generate these channel parameters.

[0099] If the relative speed between the u-th user and the base station is v u Then the maximum Doppler frequency shift is

[0100]

[0101] Where f c Let v be the carrier frequency and c be the speed of light. Then, at v... max,u Given the existing information, the Doppler frequency shifts v of all K paths can be obtained further based on the 5G 3GPP standard and the 5G toolbox provided by MATLAB. k,u .

[0102] In the 5G 3GPP standard, the TDL-A channel has a total of 23 resolvable paths, therefore K = 23;

[0103] |hk,0 (t)|,τ k,u and v k,u Both can be extended by d with a known delay. u and maximum Doppler frequency shift v max,u The channel model is obtained directly using the TDL-A channel model in MATLAB's 5GNR toolbox, and only requires the addition of a power gain p afterward. u and random phase You can get h k,u (t);

[0104] h rc [s] represents the impulse response of the raised cosine filter, which can be defined given the total number of sampled symbols s. pa and oversampling rate s ps In this case, it can be directly generated using MATLAB's rcodesign function. It can be noted that... The value is equal to the first digit of the impulse response of the discrete raised cosine filter generated by MATLAB. Values.

[0105] Assume the channel estimation error matrix ε u Let IID be a Gaussian matrix, where the variance of each term is . That is

[0106] ε u ~ i.i.d. ε u ,

[0107]

[0108] ε u This can be generated using MATLAB functions such as randn. The final estimated channel matrix is:

[0109] Generate two independent random modulation matrices respectively. and The true equivalent channel matrix and the estimated equivalent channel matrix of the u-th user are obtained as A. u =H u Ξ u ,

[0110] Assuming the signal-to-noise ratio at the receiver is snr, then the noise power is σ. 2 =1 / snr, the signal received by the base station is

[0111]

[0112] The specific process of the MU-OAMP algorithm in this embodiment includes:

[0113] Since the iteration method is similar for all users, we will only take the u-th user as an example:

[0114] Initialize the number of iterations t=1, the maximum number of iterations T; nonlinear estimator Nonlinear estimation of variance

[0115] Linear estimation (LE):

[0116] Calculate the intermediate variable z (t)

[0117]

[0118] Where ∑ (t) It can be various filter matrices, for example

[0119] 1) Matched filter:

[0120] ∑ (t) =I L ,

[0121] 2) Undevised LMMSE filter:

[0122]

[0123] 3) Modified LMMSE filter:

[0124]

[0125] in

[0126]

[0127] Calculate orthogonal parameters

[0128]

[0129] Calculate the linear estimator

[0130]

[0131] Calculate the linear estimate of variance

[0132]

[0133] In particular, when

[0134] 1)∑ (t) When it is a matched filter

[0135]

[0136] 2)∑ (t) When it is an LMMSE filter

[0137]

[0138] Nonlinear estimation (NLE):

[0139] Given the prior distribution s u ~ i.i.d. S i Calculate the posterior estimator of MMSE

[0140]

[0141] Calculate the posterior variance of the MMSE estimate

[0142]

[0143] Calculate the variance of the nonlinear estimate

[0144]

[0145] Calculate the nonlinear estimator

[0146]

[0147] To further demonstrate the practicality of this application, this embodiment includes a bit error rate calculation after each iteration:

[0148] Calculate the bit error rate (BER) for the u-th user in the current t-th iteration. u (t)

[0149]

[0150] in This is an indicator function that takes the value 1 when the condition in (·) is met, and 0 otherwise; s u,i s u The i-th element (i = 1, ..., N) u ), This represents the decoding decision result of the current t-th iteration. The i-th element. s y The value is taken from the constellation chart. That is And s u,i Take x k The probability is q k =P(s) u,i =x k ).

[0151] Specifically, for normalized QPSK signals remember So The judgment result can be obtained from the following formula.

[0152]

[0153] in

[0154]

[0155] For linear estimators The i-th element.

[0156] Because in d k,j If the value is too large or too small, MATLAB will perform calculations... Numerical overflow issues may occur, so we assign d in the actual program. k,j The upper and lower bounds are 100 and -100 respectively, denoted as We have

[0157]

[0158] The number of update iterations is t = t + 1.

[0159] Repeat the iterative process until t = T + 1.

[0160] Hard decision is made based on the obtained linear estimator and variance to obtain the final decoding result.

[0161] Furthermore, to ensure the practicality of this invention, the embodiments also provide the state evolution (SE) of the algorithm, specifically including:

[0162] The state evolution (SE) of the algorithm is as follows:

[0163]

[0164]

[0165] in:

[0166] 1. and These are the estimators of the linear and nonlinear ends in the t-th iteration, respectively.

[0167] 2. (It should be noted that for an N×N matrix U) N×N ,symbol

[0168] This indicates the search for an N×N matrix U. N×N The empirical mean of the trace, while for an N×1 one-dimensional vector r N×1 ,symbol This means finding vector r. N×1 (Empirical mean of the elements at the N positions); signal s u power

[0169] 3. Remember but

[0170] and These are the error vectors between the estimators and the true values ​​of the linear and nonlinear ends, respectively, in the t-th iteration. and

[0171] For error SE;

[0172] 4.

[0173] 5. It is a deterministic, generally non-separable function. It's important to note that these functions must satisfy the divergence-free property.

[0174]

[0175] It is important to note that Representing the Jacobi matrix

[0176]

[0177] Specifically, we can choose a function:

[0178]

[0179] in The posterior estimation function of the minimum mean square error (MMSE)

[0180]

[0181] in

[0182] 6.∑ (t) In algorithm analysis, the matrix can be almost arbitrary, but its correct construction is crucial for achieving optimal performance. The simplest choice is a matched filter:

[0183] ∑ (t) =I L ,

[0184] At this point, the OAMP algorithm iteration formula is:

[0185]

[0186] SE is

[0187]

[0188] From a computational complexity perspective, this choice is particularly attractive because it makes the aforementioned OAMP-style iterations involve only low-complexity matrix and vector multiplications. However, this simple choice can lead to poor algorithm performance.

[0189] Another commonly used option is the Linear Minimum Mean Square Error (LMMSE) estimator:

[0190]

[0191] At this point, the OAMP algorithm iteration formula is:

[0192]

[0193] SE is

[0194]

[0195]

[0196] However, when CSI is imperfect, the performance of a standard LMMSE estimator deteriorates as the CSI error increases. Therefore, to achieve optimal estimation performance under imperfect CSI conditions, the standard LMMSE estimator requires additional modifications. Specifically, we can consider the case where the error matrix is ​​a complex Gaussian matrix, i.e.

[0197]

[0198] in It is an arbitrarily fixed positive semi-definite matrix. Here, we use the notation . Let X be a matrix variable that follows a complex Gaussian distribution, with row and column covariance matrices A and B, respectively. In other words, The terms in X are IID standard normally distributed random variables. It is important to note that in the analysis we assume H... u and The spectral radius is confined to a certain limit with a very high probability. Formally, we assume a normalized scaling factor:

[0199]

[0200] In the formula, ||·||² represents the maximum singular value of the matrix. Note that if for some constant c > 0, We can always achieve this by performing an L-factor on the observation vector. -c Scaling by times to make Therefore, we assume And that is why, ε u The scale of the row / column covariance matrix in the model is:

[0201] Furthermore, it can be proven that under the Gaussian channel estimation error model

[0202]

[0203] It can be noted that under this error model, the OAMP algorithm has the following characteristics:

[0204] 1) The posterior estimation function of MMSE is:

[0205]

[0206] 2) The nonlinear estimation function is

[0207]

[0208] 3) SE can be simplified to

[0209]

[0210]

[0211] in

[0212]

[0213] At this point, we will use the ∑ in the LMMSE estimator (t) Revised to:

[0214]

[0215] The OAMP algorithm iteration formula using the modified LMMSE estimator is now as follows:

[0216]

[0217] SE can be further simplified to:

[0218]

[0219] Furthermore, it can make At this point, the error matrix is ​​an IID Gaussian matrix with variance of... That is

[0220] ε u ~ i.i.d. ε u ,

[0221] in

[0222]

[0223] at this time

[0224]

[0225] according to Figure 3 The diagram illustrates the changes in the system bit error rate (BER) with the number of iterations in a TDL-A channel model, with U = 2 users, accurate channel estimation (σ1 = σ2 = 0), a QPSK signal, L = 4096 subcarriers, and transmit signal lengths of users 1 and 2 (N1 = 4096, N2 = 4096 and 13dB respectively). Users 1 and 2 move at a speed of 100km / h relative to the base station, with a delay spread of 10ns. In an uplink communication scenario where users 1 and 2 have power gains of 0dB and -3dB respectively, the diagram shows the BER changes with the number of iterations when using the standard OAMP algorithm and the multi-user OAMP algorithm for signal recovery. The dashed and solid lines represent the BER curves for the standard OAMP algorithm and the multi-user OAMP algorithm, respectively. The red and blue lines represent the BER curves for users 1 and 2, respectively. The red and blue circles represent the signal strength (SE) of users 1 and 2 when using the multi-user OAMP algorithm for signal recovery.

[0226] according to Figure 4The diagram illustrates the changes in the system bit error rate (BER) with the number of iterations in a TDL-A channel model, with U = 2 users, accurate channel estimation (σ1 = σ2 = 0), a QPSK signal, L = 4096 subcarriers, and transmit signal lengths of users 1 and 2 (N1 = 4096, N2 = 2048, respectively), with a signal-to-noise ratio (SNR) of snr = 13 dB. Users 1 and 2 move at a speed of 100 km / h relative to the base station, and both have a delay spread of 10 ns. In an uplink communication scenario where users 1 and 2 have power gains of 0 dB and -3 dB respectively, the diagram shows the BER changes with the number of iterations when using the standard OAMP algorithm and the multi-user OAMP algorithm for signal recovery. The dashed and solid lines represent the BER curves for the standard OAMP algorithm and the multi-user OAMP algorithm, respectively; the red and blue lines represent the BER curves for users 1 and 2, respectively; and the red and blue circles represent the signal strength (SE) for users 1 and 2 when using the multi-user OAMP algorithm for signal recovery.

[0227] according to Figure 5 The figure shows the variation of the system bit error rate (BER) with the number of iterations in the following scenarios: a TDL-A channel model, U=2 users, channel estimation error matrices of IID Gaussian matrices with error parameters σ1=0.135 and σ2=0.096, QPSK signals, L=4096 subcarriers, and transmit signal lengths of users 1 and 2 of N1=4096 and N2=4096, respectively, with a signal-to-noise ratio of snr=18dB; the relative speeds of users 1 and 2 to the base station are both 100km / h, and the delay spread is 10ns; and the uplink communication scenarios are 0dB and -3dB for users 1 and 2, respectively. These scenarios involve using an undevised LMMSE filter and a modified LMMSE filter for signal recovery. The dashed and solid lines represent the bit error rate curves using the uncorrected and corrected LMMSE filters, respectively. The red and blue lines represent the bit error rate curves for user 1 and user 2, respectively. The red and blue circles represent the signal error rate (SE) for user 1 and user 2 when using the corrected LMMSE filter for signal recovery.

[0228] according to Figure 6The figure shows the changes in the system bit error rate (BER) with the number of iterations in the following scenarios: a TDL-A channel model, U=2 users, channel estimation error matrices of IID Gaussian matrices with error parameters σ1=0.135 and σ2=0.096, QPSK signals, L=4096 subcarriers, and transmit signal lengths of users 1 and 2 of N1=4096 and N2=4096, respectively, with a signal-to-noise ratio of snr=18dB; the relative speeds of users 1 and 2 to the base station are both 100km / h, and the delay spread is 10ns; and the uplink communication scenarios are 0dB and -3dB for users 1 and 2, respectively. These scenarios involve signal recovery using a matched filter and a modified LMMSE filter, respectively. The dashed and solid lines represent the bit error rate curves using the matched filter and the modified LMMSE filter, respectively. The red and blue lines represent the bit error rate curves for user 1 and user 2, respectively. The red and blue circles represent the signal error rate (SE) for user 1 and user 2 when using the modified LMMSE filter for signal recovery.

[0229] according to Figure 7 The figure shows the variation of the system bit error rate (BER) with the number of iterations in the following scenarios: a TDL-A channel model, U=2 users, channel estimation error matrices of IID Gaussian matrices with error parameters σ1=0.250 and σ2=0.125, QPSK signal, L=4096 subcarriers, and signal lengths of users 1 and 2 of N1=4096 and N2=2048 respectively, with a signal-to-noise ratio of snr=18dB; the relative speed of users 1 and 2 to the base station is 100km / h, and the delay spread is 10ns; and the uplink communication scenarios with power gains of 0dB and -6dB for users 1 and 2 respectively, using an undevised LMMSE filter and a modified LMMSE filter for signal recovery. The dashed and solid lines represent the bit error rate curves using the uncorrected and corrected LMMSE filters, respectively. The red and blue lines represent the bit error rate curves for user 1 and user 2, respectively. The red and blue circles represent the signal error rate (SE) for user 1 and user 2 when using the corrected LMMSE filter for signal recovery.

[0230] according to Figure 8The figure shows the variation of the system bit error rate (BER) with the number of iterations in the following scenarios: a TDL-A channel model, U=2 users, channel estimation error matrices of IID Gaussian matrices with error parameters σ1=0.250 and σ2=0.125, QPSK signal, L=4096 subcarriers, and signal lengths of users 1 and 2 of N1=4096 and N2=2048 respectively, with a signal-to-noise ratio of snr=18dB; the relative speed of users 1 and 2 to the base station is 100km / h, and the delay spread is 10ns; and the uplink communication scenarios of users 1 and 2 of 0dB and -6dB respectively, using a matched filter and a modified LMMSE filter for signal recovery. The dashed and solid lines represent the bit error rate curves using the matched filter and the modified LMMSE filter, respectively. The red and blue lines represent the bit error rate curves for user 1 and user 2, respectively. The red and blue circles represent the signal error rate (SE) for user 1 and user 2 when using the modified LMMSE filter for signal recovery.

[0231] Example 3:

[0232] Specifically, embodiments of the present invention also provide a SISO time-varying (including static) multipath channel model used in simulating the transmission and reception processes of the transmitting and receiving ends:

[0233] We have considered the Time Delay Line (TDL) channel in 5G 3GPP, and the channel model is the same as in Example 2. l = 0, 1, ..., L-1;

[0234] This invention proposes an OAMP algorithm for signal detection in uplink multi-user communication scenarios. This algorithm can still have good detection performance even when there are significant differences in the CSI obtained by each user through channel estimation and when there may be errors between the CSI obtained by each user through channel estimation and their respective true CSI.

[0235] This embodiment is an example of multi-user OAMP detection with signal spread and random modulation under static channel conditions (i.e., channel Doppler shift is 0) and when the channel estimation is perfect. The implementation of this embodiment is as follows: the communication channel is a static multipath channel, and the CSI obtained by each user through channel estimation is different and may have obvious differences, but the estimated channel is exactly the same as the real channel.

[0236] The specific steps of multi-user OAMP include:

[0237] S1. Initialization:

[0238] S11. For the sake of brevity, assume that both User 1 and User 2 use the TDL-A channel in the 5G 3GPP standard (but in reality, the channel categories of the two users can be completely different).

[0239] S111. Set the number of subcarriers to L, the subcarrier spacing to Δf, and the carrier frequency to f. c Therefore, the sampling frequency f can be calculated. s =LΔf, sampling interval T s =1 / f s Set the total number of sampling symbols to s. pa The oversampling rate is s ps .

[0240] S112. Assume that there is no relative motion between user 1 and user 2 and the base station, that is, the Doppler frequency shift is 0.

[0241] S113. Assume that the power gains of user 1 and user 2 due to slow fading of the channel are p1 and p2, respectively.

[0242] S114, the delay spreads of User 1 and User 2 are d1 and d2, respectively.

[0243] S12,

[0244] S121. Generate the channel matrix for the u-th user (u = 1, 2) based on the following channel model.

[0245]

[0246] in

[0247]

[0248] The parameters involved in the above channel model are explained in Example 1.

[0249] S13. Assume the base station receives QPSK signals transmitted by two users (U=2), and the transmission signal lengths of user 1 and user 2 are N1 and N2, respectively; let the transmission signals of user 1 and user 2 be N1 and N2, respectively. and Then there is

[0250] s1~ i.i.d. S1,

[0251] s2~ i.i.d. S2,

[0252] in

[0253]

[0254] u = 1, 2,

[0255] I u Q u This can be generated using MATLAB's built-in uniform distribution function, randi. The signal amplitude here is not... The reason is that the signal needs to be padded with zeros and extended so that the total signal length matches the number of columns L of the channel matrix.

[0256]

[0257] To ensure that the average power of the signal remains at 1, the amplitude of the signal needs to be adjusted.

[0258] S14. Perform Fast Fourier Transform (FFT) on both users respectively (equivalent to multiplying by a Fast Fourier Transform matrix F), then perform independent random interleaving between the users (equivalent to multiplying by a random permutation matrix ∏1 or ∏2), and finally perform an Inverse Fourier Transform (equivalent to multiplying by a Fast Inverse Fourier Transform matrix F). H The resulting random modulated signals belonging to the two users are then fed into static channels H1 and H2, respectively. Assuming the signal-to-noise ratio at the receiver is snr, the noise power is σ. 2 =1 / snr, the signal received by the base station is

[0259]

[0260] Finally, the equivalent channel matrices obtained by performing Fourier transform on the received signal at the receiving end are A1 = A1Π1F and A2 = A2Π2F, respectively. Where Λ1 = FH1F H Λ2=FH2F H They are two diagonal matrices, that is,

[0261]

[0262] S2, MU-OAMP algorithm

[0263] Since the MU-OAMP algorithm used here is exactly the same as that in Example 1, it will not be described again.

[0264] In addition to the channel model and MU-OAMP algorithm principle already mentioned in the technical principles of Embodiment 1, the technical principles of the OAMP algorithm for uplink multi-user static communication scenario signal detection, as described in the above embodiment, also include:

[0265] In static communication scenarios, the channel is a static multipath channel (without Doppler shift). By adding a cyclic prefix (CP) at the transmitter, the static multipath channel matrix can be diagonalized by left-multiplying by the Fourier transform matrix and right-multiplying by the inverse Fourier transform matrix.

[0266] When using multi-user OAMP detection in static scenes, we can let the random modulation matrix Ξ u It is a square matrix as follows

[0267] Ξ u =F H Π u F,

[0268] in, It's an interleaver. After performing a Fourier transform on the received data, the equivalent channel becomes...

[0269] A u =Λ u Π u F.

[0270] Therefore, by using the aforementioned random modulation matrix, the time complexity of the multi-user OAMP algorithm with LMMSE filters can be significantly reduced. This is because the matrix inversion operation is performed on the diagonal matrix, thus reducing the time complexity from... Reduce to Furthermore, due to the presence of random interleaved FFT operations, the time complexity required for matrix-vector multiplication can be reduced from... Down to

[0271] The communication scenarios discussed so far have all been dynamic, featuring time-varying multipath channels. Due to the Doppler shift, the commonly used Orthogonal Frequency Division Multiple Access (OFDMA) technology cannot diagonalize the channel matrix, thus preventing direct application of MMSE for signal detection. However, in static scenarios, we can compare the detection performance of multi-user OAMP detection with random modulation and OFDMA-based MMSE detection. For simplicity, we can assume each user has perfect CSI. Simultaneously, to ensure fairness, we should maintain the same amount of information carried by the signal as with OFDMA technology under the same channel size. Therefore, we extend the signal without changing the total amount of information carried, which is equivalent to padding the transmitted signal with zeros.

[0272]

[0273] Where s u The length (i.e., N) u The power and signal strength should be compatible with those used for signal transmission with OFDMA technology. u Consistent.

[0274] It should be noted that in a dual-user static scenario, the principle of OFDMA technology is as follows:

[0275]

[0276] in

[0277]

[0278] like Figure 9 The following is an example of the following scenarios: In an uplink communication scenario with a TDL-A channel model, U=2 users, accurate channel estimation (σ1=σ2=0), QPSK signal transmission, L=4096 subcarriers, and transmission signal lengths of users 1 and 2 (N1=N2=2048 respectively); users 1 and 2 are stationary relative to the base station, with a delay spread of 10ns for both; and power gains of users 1 and 2 are 0dB and -3dB respectively. The results show the variation of the system bit error rate (BER) with signal-to-noise ratio (SNR) when performing (i) multi-user OAMP detection with signal spread and random modulation, and (ii) MMSE detection using OFDMA technology for signal modulation. The solid and dashed lines represent the bit error rate curves of (i) multi-user OAMP detection with signal spread and random modulation, and (ii) MMSE detection with signal modulation using OFDMA technology, respectively. The red and blue lines represent the bit error rate curves of user 1 and user 2, respectively. The red and blue circles represent the SE of user 1 and user 2 when using multi-user OAMP detection with signal spread and random modulation, respectively.

[0279] In summary, the method of the present invention can achieve detection performance significantly better than MMSE detection based on OFDMA technology under static channel conditions.

[0280] Corresponding to the aforementioned embodiment of an uplink multi-user random modulation and OAMP detection method, the present invention also provides an embodiment of an uplink multi-user random modulation and OAMP detection device.

[0281] See Figure 10 The present invention provides an uplink multi-user random modulation and OAMP detection device, including a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it is used to implement an uplink multi-user random modulation and OAMP detection method in the above embodiment.

[0282] An embodiment of the uplink multi-user random modulation and OAMP detection device provided by this invention can be applied to any device with data processing capabilities, such as a computer. The device embodiment can be implemented in software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 10 The diagram shown is a hardware structure diagram of any data processing-capable device where an uplink multi-user random modulation and OAMP detection device provided by the present invention is located. Except for... Figure 10 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0283] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0284] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0285] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements an uplink multi-user random modulation and OAMP detection method as described in the above embodiments.

[0286] The computer-readable storage medium can be an internal storage unit of any data processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device of any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of any data processing device. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0287] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the aforementioned uplink multi-user random modulation and OAMP detection method.

[0288] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0289] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application. This application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for uplink multi-user random modulation and OAMP detection, characterized in that, include: S1. Transmit the signal after performing independent random modulation on each user at the transmitting end; S2. Receive the sum of signals from each user at the receiving end; S3. Based on the received signal and the state information of the channel matrix, the estimation error of the channel matrix and the channel noise power, perform linear estimation; The orthogonal parameters are calculated based on the channel state information, the estimation error of the channel matrix, and the channel noise power. The linear estimator for the current iteration is calculated based on the orthogonal parameters, channel state information, estimation error of the channel matrix, channel noise power, and nonlinear estimator from the previous iteration. The linear estimation variance in this iteration is obtained based on the orthogonal parameters, channel state information, estimation error of the channel matrix, channel noise power, and nonlinear estimation variance in the previous iteration. S4. Based on the prior distribution of each user's transmitted signal, the estimator and variance obtained by linear estimation, nonlinear estimation is performed to obtain the nonlinear estimator and variance in this iteration; The linear and nonlinear estimates are iterated continuously until convergence or the maximum number of iterations is reached, and the linear estimator and variance of the last iteration are obtained. S5. Make a hard decision based on the obtained linear estimator and variance to obtain the final decoding result.

2. The uplink multi-user random modulation and OAMP detection method according to claim 1, characterized in that, The transmission of signals after independent random modulation for each user specifically includes: x u =Ξ u s u , And the equivalent channel matrix A u =H u Ξ u It belongs to the universal class of matrices; among which It is of dimension L×N u A random unitary matrix, It is of dimension N u A ×1 unmodulated input signal vector; This represents the multiple transmit multiple receive channel matrix corresponding to the u-th user.

3. The uplink multi-user random modulation and OAMP detection method according to claim 1, characterized in that, The receipt of the sum of signals from each user from the receiving end includes: in It is the sum of the signals received by the base station from U users after passing through their respective channels and being affected by noise interference; It is the Gaussian channel noise vector of IID, with It is the transmitted signal vector from the u-th user. This represents the multiple transmit multiple receive channel matrix corresponding to the u-th user, where dimension L is the number of subcarriers and signal x u It is a signal that has been randomly modulated.

4. The uplink multi-user random modulation and OAMP detection method according to claim 1, characterized in that, The linear estimation includes: Calculate the intermediate variable z (t) Where Σ (t) is the filter matrix; y is the sum of the received multi-user signals. To estimate the equivalent channel matrix, Here is the nonlinear estimate for the u-th user, where U is the total number of users; the estimated channel matrix is... Calculate orthogonal parameters Calculate the linear estimator Calculate the linear estimate of variance in, It is of dimension N u An unmodulated input signal vector of 1 × 1.

5. The uplink multi-user random modulation and OAMP detection method according to claim 4, characterized in that, The filter matrix Σ (t) When it is a matched filter, the variance of the linear estimate is: Where I L Let represent an identity matrix of dimension L×L, where L is the number of subcarriers; the subscript u′ represents the u′-th user, u′=1,…,U, and the filter matrix ∑ (t) When using an LMMSE filter, the variance of the linear estimate is:

6. The uplink multi-user random modulation and OAMP detection method according to claim 4, characterized in that, The nonlinear estimation includes: Given the prior distribution s u ~ i.i.d. S u Calculate the posterior estimator of MMSE Calculate the posterior variance of the MMSE estimate Calculate the variance of the nonlinear estimate Calculate the nonlinear estimator S24, The number of update iterations is t = t + 1.

7. The uplink multi-user random modulation and OAMP detection method according to claim 1, characterized in that, When the channels at both the transmitter and receiver are static channels, the random modulation process involves performing a Fast Fourier Transform (FFT) on both users, followed by independent random interleaving between the users, and finally an Inverse Fourier Transform (IFT) to obtain the random modulated signal. The random modulation matrix Ξ is then used. u As shown below: X u =F H P u F, in, It is an interleaver, and F is the Fourier transform.

8. The uplink multi-user random modulation and OAMP detection method according to claim 6, characterized in that, The hard decision-making process based on the obtained linear estimator and variance specifically includes: For normalized QPSK signals remember So The judgment result is obtained from the following formula. in This represents the i-th element of the decoding decision result. For the linear estimator in the last iteration The i-th element; Give d k,q The upper and lower bounds are set to 100 and -100 respectively.

9. An uplink multi-user random modulation and OAMP detection device, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the processor executes the executable code, it implements an uplink multi-user random modulation and OAMP detection method as described in any one of claims 1-8.

10. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements an uplink multi-user random modulation and OAMP detection method as described in any one of claims 1-8.

Images