Method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition

Abstract

The invention provides a method for realizing millimeter wave MIMO (Multiple Input Multiple Output) system channel estimation based on random tensor network decomposition, which comprises the following steps: k subcarriers and a transmitting signal on a tth time frame are acquired, a transmitting antenna transmits the transmitting signal to a receiving antenna, and the transmitting signal is expressed as xk (t) = sk (t) ck (t), # imgabs0 #, sk (t) represents an orthogonal pilot signal on the kth subcarrier, Ck (t) represents a precoding matrix on the kth subcarrier, and k subcarriers are represented as x (t) = [x1 (t),..., xK (t)] T on a time frame; through the random grid tensor decomposition method, complex signal processing in the millimeter wave large-scale MIMO antenna array is solved, accurate estimation of channel state information and wireless data transmission with cooperation of multiple communication modes are achieved, the flexibility and reliability of the system are improved, and the requirements of different application scenes are met. The adaptive network adaptation and service are provided for the user, the user experience is improved, and the high-quality communication service can be obtained in different network environments.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and more specifically, to a method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition. Background Art

[0002] Currently, the digital economy is an important direction for economic transformation and upgrading. With the booming era of new technologies, new business forms, and new platforms such as 5G and artificial intelligence, digital industrialization and industrial digitalization are promoted, and digital technologies are deeply integrated with economic and social development.

[0003] In recent years, due to the increasing demand for service rates faced by microwave communications, there are great challenges due to limited spectrum bandwidth in traditional frequency bands. In the millimeter-wave E-band frequency band, there is approximately 10 GHz of spectrum available, and it is easy to achieve high-speed microwave transmission of Gbps and 10 Gbps rate levels for point-to-point (or point-to-multipoint).

[0004] The key indicators of 5G mobile communication systems include ultra-high traffic density (exceeding 10 Mbit / s / Km2), ultra-high transmission rate (peak rate greater than 10 Gbit / s), lower network latency (air interface latency less than 1 ms), etc.

[0005] In existing wireless communication systems, the spectrum resources in the low frequency band (<6 GHz) are tense, while there are 45 GHz of unused spectrum resources in the frequency range of 6 GHz - 100 GHz, which can meet the future 5G requirements for higher system capacity and Gbps transmission rate. Compared with low-frequency communication systems, the millimeter-wave band can be allocated a wider bandwidth, thus greatly improving the channel capacity; at the same time, the millimeter-wave signal has a shorter wavelength, and a larger-scale antenna array can be configured within the same physical size at the base station or user end, and large-scale MIMO technology can be effectively used to achieve flexible multi-user intelligent beamforming, and the spectrum efficiency can be greatly improved through spatial multiplexing.

[0006] At the same time, due to the high directivity of millimeter-wave large-scale arrays and the large free-space propagation loss in the millimeter-wave band, millimeter-wave communication is more difficult to intercept and interfere, thus ensuring communication security and confidentiality.

[0007] Currently, there are still some problems and challenges in the application of millimeter-wave technology in 5G mobile communication systems, such as millimeter-wave chips with small size, high integration, and low power consumption (multi-channel transceiver chips, high-frequency frequency sources, etc.); low-loss millimeter-wave large-scale antenna arrays; on-chip antennas for package-level (SiP) systems; millimeter-wave propagation characteristics and channel modeling for different application scenarios; highly integrated and integrated millimeter-wave large-scale MIMO radio frequency hardware systems, etc.

[0008] Using a millimeter-wave massive antenna array (MIMO), the millimeter-wave signal can be propagated farther from the base station.

[0009] Since the antenna size is proportional to the signal wavelength, a large-scale MIMO antenna array can be deployed in a millimeter-wave system without occupying too much space. At the same time, under a massive MIMO array, the signal energy can be concentrated in an extremely narrow beam in space and precisely directed to the downlink user, thereby maximizing the propagation distance in this direction.

[0010] A large-scale MIMO antenna array can make full use of the spatial multiplexing ability of multiple antennas, greatly improving the overall spectral efficiency of the system and overcoming severe path loss.

[0011] At the same time, in 5G millimeter-wave high-speed mobile communication, ultra-high mobility, low latency, ultra-dense user distribution, and channel time-variation will lead to Doppler frequency shift, high penetration loss, increased power consumption during frequent handover processes, reduced transmission data rate, and relatively high computational complexity of traditional channel estimation schemes. Summary of the Invention

[0012] The technical problem to be solved by the present invention is to provide a millimeter-wave MIMO system channel estimation method based on random tensor network decomposition that uses a random grid tensor decomposition algorithm, avoids iterative and random initialization operations, and effectively reduces computational complexity, in view of the deficiencies in the above technical solutions.

[0013] The present invention provides a millimeter-wave MIMO system channel estimation method based on random tensor network decomposition. In the millimeter-wave MIMO system, the transmitting end is equipped with Nt transmitting antennas, the receiving end is equipped with Nr receiving antennas, and the millimeter-wave MIMO system has a total of k subcarriers, where k subcarriers are used for channel estimation. The method includes the following steps:

[0014] S1. Obtain the transmitted signal on k subcarriers and the t-th time frame. The transmitting antenna sends the transmitted signal to the receiving antenna. The transmitted signal is expressed as x k (t) = s k (t)c k (t), where, S k (t) represents the orthogonal pilot signal on the k-th subcarrier, C k (t) represents the precoding matrix on the k-th subcarrier. Express the k subcarriers on the time frame as x(t) = [x 1 (t),..., x K (t)] T ;

[0015] S2. Utilizing the sparse scattering characteristics of the millimeter-wave channel and the spatial structure of the tensor, the receiving antenna obtains the received signal, stacks the received signal into a three-dimensional tensor form, and constructs three factor matrices corresponding to the three-dimensional tensor form. The received signal includes the number of users, receiving antennas, subcarriers, and pilot symbols. The random grid tensor decomposition algorithm is used to estimate the channel parameters of the first three time slots of the time-varying channel for the channels between the receiving antenna and subcarriers, between the receiving antenna and pilot symbols, and between the receiving antenna and users.

[0016] S3. Utilize the GOLAY sequences Ga N and Gb N to estimate the factor matrices in the time domain respectively, and obtain the estimated channel results.

[0017] In the method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition according to the present invention; the step S1 includes the following steps:

[0018] S11. There are P multipaths set between the transmitting end and the receiving end, and the time delay caused by each multipath is τ p , then the spatio-temporal domain channel is expressed as where α p represents the attenuation coefficient on the P-th path, θ p ∈[0, 2π] and respectively represent the angle of arrival and the angle of departure on the P-th path, δ(·) represents the impulse response function, τ p represents the delay on the P-th path, and α r (·) and α t (·) respectively represent the uniform antenna array responses of the receiving end and the transmitting end.

[0019] In the method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition according to the present invention; the step S1 further includes the following steps:

[0020] S12. The receiving end preprocesses the received signal obtained by the receiving antenna, and uses a correlator to receive each multipath received signal. At the receiving end, the j-th receiving antenna's P-th multipath received signal is preprocessed, and the preprocessed multipath received signal is output as Z j，p (t) = ∫r j (t)w(t - τ j，p )dt, where τ j，p represents the correlation time delay of the j-th receiving antenna's p-th multipath signal, w(t - τ) represents the synchronous correlation function corresponding to the multipath time delay; Rj(t) represents the j-th received signal.

[0021] In the millimeter-wave MIMO system channel estimation method based on random tensor network decomposition according to the present invention; the step S1 further includes the following steps:

[0022] S13, the transmitting end preprocesses the transmitted signal of the transmitting antenna to obtain a transmitted signal Z j [n] makes a decision on the transmitted signal Z j [n] to obtain the finally output transmitted signal Y j , pre-encodes the transmitted signal Y j into a matrix form Among them, is expressed as a matrix expansion formula expanded along the receiving antenna latitude, is expressed as a transmitted signal matrix formed on the Nt-th transmitting antenna at the K-th subcarrier, SD J (H) represents a diagonalization operation, and the j-th row vector H of the channel transmission matrix H is taken out J , using the elements of H J as the main diagonal elements, and other elements are recorded as zeros, is expressed as a noise matrix, and its elements are independent and identically distributed additive white Gaussian noises, is expressed as a pre-coding matrix for the transmitted signal, and R represents a non-negative real matrix operation.

[0023] In the millimeter-wave MIMO system channel estimation method based on random tensor network decomposition according to the present invention; the step S2 includes the following steps:

[0024] S21, stack the received signals of Nr receiving antennas into a three-dimensional tensor form Y, where,

[0025]

[0026] ⊙ represents the Khatri-Rao product, H represents a row vector, SD J (H) represents a diagonalization operation.

[0027] In the millimeter-wave MIMO system channel estimation method based on random tensor network decomposition according to the present invention; the step S2 further includes the following steps:

[0028] S22, construct The corresponding three factor matrices are Among them, respectively represent the column vectors of the factor matrices H, S, and C, and Ν represents the corresponding noise tensor.

[0029] In the millimeter-wave MIMO system channel estimation method based on random tensor network decomposition according to the present invention; the step S2 further includes the following steps:

[0030] S23. Extract random vectors ω1, ω2, ω3 as the approximate basis of the n-order tensor model through Gaussian distribution, and form a measurement matrix with the random vectors The column space for the user channel tensor slice Ωn is denoted as S n = y (n) Ω n , n, 1, 2, …, N, where y(n) represents the n-mode unfolding of the tensor y; Denote it as the channel sample matrix, and use the sample matrix as the approximate basis of the n-mode unfolding tensor;

[0031] S24. Sample the n-mode unfolding tensor y(n) of the received signal through the random projection method, and decompose the channel sample matrix Sn into: S n = Q n R n .; Construct an equivalent channel sample matrix using the Schmidt orthogonal method:

[0032] And through random tensor decomposition, obtain the received signal of each user, which is represented by a third-order tensor as where E i represents the error objective function under the Euclidean distance, C represents the set of complex numbers, where each mode of the tensor represents the receive antennas (Nr), subcarriers (K), and symbols (T) respectively.

[0033] S25. Assume that the transmitter sends the same pilot symbol, and let The unfolding form of one receive antenna mode in the received signal tensor u is:

[0034]

[0035] where B, G, A are three factor matrices; N is the noise matrix; is the attenuation coefficient matrix of all Nr receive antennas; @ represents the KhatriRao product, and we get: u = I p × 1 B × 2 G × 3 A + N, let A be one receive antenna mode in the received signal tensor u; then 1B, 2G, 3A represent 3 factor matrices;

[0036] where Ip is the first-order identity tensor; N is the corresponding noise tensor; the symbol Xn represents the multiplication of the n-mode of the tensor and the matrix;

[0037] S26. The input tensor is The input relative error is ε, and the output satisfies And H, S, C are non - negative matrices. Initialize the non - negative matrices H, S, C randomly. Fix S and C, and update H as follows:

[0038]

[0039] Fix H and C, and update S as follows:

[0040] Fix H and S, and update C as follows: Repeat the loop (1), (2), (3) until the relative error is less than ε, then stop the iteration.

[0041] In the method for channel estimation of millimeter - wave MIMO systems based on random tensor network decomposition according to the present invention; step S3 includes the following steps:

[0042] S31, at the transmitter end, r a (k) and r b (k) are a pair of Gray sequences Ga N and Gb N with a length of N points for transmission. Through the multipath channel and additive white Gaussian noise, the received sequences are respectively: Where h(k) represents the actual multipath channel impulse response, and “*” represents the convolution operation. n a (k) and n b (k) represent Gaussian noise with a mean of 0 and a variance of σ 2 .

[0043] In the method for channel estimation of millimeter - wave MIMO systems based on random tensor network decomposition according to the present invention; step S3 further includes the following steps:

[0044] S32, Correlate the received r a (k) and r b (k) with the local Gray sequences Ga N and Gb N respectively, and add the results of the cross - correlation operations to obtain the channel impulse response r CIR (k), where Where Ra N (k) and Rb N represent the autocorrelation results of the Gray sequences Ga N and Gb N respectively. n(k) represents Gaussian noise with a mean of 0 and a variance of 2Nσ 2 . At this time, the length of the channel impulse response r CIR (k) is twice the length of the Gray sequence Ga N .

[0045] In the method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition according to the present invention; step S3 further includes the following steps:

[0046] S33, calculate the channel power of the channel impulse response r CIR (k), calculate the average power of the first half of the channel impulse response r CIR (k), this average power is the noise power. Divide the calculated signal power by the estimated noise power to obtain the signal-to-noise ratio, and amplify the channel power by 4N 2 times. At this time, the corresponding increase multiple of the signal-to-noise ratio is 2N.

[0047] The method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition according to the present invention solves the complex signal processing in a millimeter-wave large-scale MIMO antenna array through the random grid tensor decomposition method, realizes the accurate estimation of channel state information and the wireless data transmission with coordinated multiple communication modes, improves the flexibility and reliability of the system, and meets the requirements of different application scenarios. It provides adaptive network adaptation and services for users, enhances the user experience, and ensures high-quality communication services in different network environments. Description of the Drawings

[0048] Figure 1 is a schematic flowchart of an embodiment of the method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition according to the present invention;

[0049] Figure 2 is a schematic diagram of the random grid tensor decomposition algorithm in the method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition according to the present invention;

[0050] Figure 3 is a schematic diagram of the channel result of GOLAY sequence estimation in the method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition according to the present invention;

[0051] Figure 4 is a comparison diagram of the estimated mean square errors of different algorithms in the method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition according to the present invention;

[0052] Figure 5 is a comparison of the bit error rates of different algorithms at different signal-to-noise ratios in the method for channel estimation of a millimeter-wave MIMO system based on random tensor network decomposition according to the present invention. Detailed Embodiments

[0053] To make the objectives, technical solutions and advantages of the present invention more comprehensible, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0054] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order different from those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0055] As Figures 1 - 3 shown, it is a schematic flowchart of an embodiment of a method for channel estimation in a millimeter-wave MIMO system based on random tensor network decomposition. A method for channel estimation in a millimeter-wave MIMO system based on random tensor network decomposition is provided. In the millimeter-wave MIMO system, the transmitting end is equipped with Nt transmitting antennas, and the receiving end is equipped with Nr receiving antennas. The millimeter-wave MIMO system has a total of k subcarriers, and k subcarriers are used for channel estimation, including the following steps:

[0056] In step S1, obtain the k subcarriers and the transmitted signal in the t-th time frame. The transmitting antenna sends the transmitted signal to the receiving antenna. The transmitted signal is expressed as x k (t)=s k (t)c k (t), where S k (t) represents the orthogonal pilot signal on the k-th subcarrier, and C k (t) represents the precoding matrix on the k-th subcarrier. The k subcarriers are expressed as x(t)=[x 1 (t),..., x K (t)] T ;

[0057] In step S2, by utilizing the sparse scattering characteristics of the millimeter-wave channel and the spatial structure of the tensor, the receiving antenna obtains a received signal, stacks the received signal into a three-dimensional tensor form, and constructs three factor matrices corresponding to the three-dimensional tensor form. The received signal includes the number of users, receiving antennas, subcarriers, and pilot symbols. The random grid tensor decomposition algorithm is used to estimate the channel parameters of the first three time slots of the time-varying channel for the channels between the receiving antenna and subcarriers, the receiving antenna and pilot symbols, and the receiving antenna and users.

[0058] In step S3, using the GOLAY sequences Ga N and Gb N respectively estimate the factor matrices in the time domain to obtain the estimated channel results.

[0059] In this embodiment, step S1 includes the following steps:

[0060] In step S11, P multipaths are set between the transmitting end and the receiving end, and the time delay caused by each multipath is τ p , then the spatio-temporal domain channel is expressed as where α p represents the attenuation coefficient on the P-th path, θ p ∩[0, 2π] and respectively represent the angle of arrival and the angle of departure on the P-th path, δ(·) represents the impulse response function, τ p represents the delay on the P-th path, α r (·) and α t (·) respectively represent the uniform antenna array responses of the receiving end and the transmitting end.

[0061] In this embodiment, step S1 further includes the following steps:

[0062] In step S12, the receiving end preprocesses the received signal obtained by the receiving antenna, and uses a correlator to receive each multipath received signal. At the receiving end, the j-th receiving antenna's P-th multipath received signal is preprocessed, and the preprocessed multipath received signal is output as Z j，p (t) = ∫r j (t)w(t - τ j，p )dt, where τ j，p represents the correlation time delay of the j-th receiving antenna's p-th multipath signal, w(t - τ) represents the synchronous correlation function corresponding to the multipath time delay; Rj(t) represents the j-th received signal.

[0063] In this embodiment, step S1 further includes the following steps:

[0064] In step S13, the transmitting end preprocesses the transmission signal of the transmitting antenna to obtain the transmission signal Z j [n], and makes a decision on the transmission signal Z j [n] to obtain the finally output transmission signal Y j , and pre-encodes the transmission signal Y j into a matrix form wherein, is expressed as a matrix expansion formula unfolded along the receiving antenna latitude, is expressed as a transmission signal matrix formed on the Nt-th transmitting antenna at the K-th subcarrier, SD J (H) represents a diagonalization operation, and the j-th row vector H of the channel transmission matrix H is taken out J , and the elements of H J are used as the main diagonal elements, and the other elements are recorded as zeros, is expressed as a noise matrix, and its elements are independent and identically distributed additive white Gaussian noise, is expressed as a pre-coding matrix for the transmission signal, and R represents a non-negative real number matrix operation.

[0065] In this embodiment, step S2 includes the following steps:

[0066] In step S21, the received signals of Nr receiving antennas are stacked into a three-dimensional tensor form Y, wherein,

[0067]

[0068] ⊙ represents the Khatri-Rao product, H represents a row vector, SD J (H) represents a diagonalization operation.

[0069] In this embodiment, step S2 further includes the following steps:

[0070] In step S22, construct The corresponding three factor matrices are wherein, respectively represent the column vectors of the factor matrices H, S, C, and Ν represents the corresponding noise tensor.

[0071] In this embodiment, step S2 further includes the following steps:

[0072] In step S23, random vectors ω1, ω2, ω3 are extracted through the Gaussian distribution as the approximate basis of the n-order tensor model, and the random vectors are formed into a measurement matrix used to represent the column space of the user channel tensor slice Ωn as S n = y (n) Ω n, n = 1, 2, …, N, where y(n) represents the n-th mode expansion of the tensor y; denoted as the channel sample matrix, and the sample matrix is used as an approximate basis for the mode-n expansion tensor;

[0073] In step S24, the received signal mode-n expansion tensor y(n) is sampled by the random projection method, and the channel sample matrix Sn is decomposed into: S n = Q n R n .; The Schmidt orthogonal method is used to construct an equivalent channel sample matrix:

[0074] And through random tensor decomposition, the received signal of each user is obtained, which is represented by a third-order tensor as where E i represents the error objective function under the Euclidean distance, C represents the set of complex numbers, where each mode of the tensor represents the receiving antenna (Nr), subcarrier (K), and symbol (T), respectively.

[0075] In step S25, assuming that the transmitter sends the same pilot symbol, let The expansion form of one receiving antenna mode in the received signal tensor u is:

[0076]

[0077] where B, G, and A are three factor matrices; N is the noise matrix; is the attenuation coefficient matrix of all Nr receiving antennas; @ represents the Khatri-Rao product, and we get: u = I p × 1 B × 2 G × 3 A + N. Let A be one receiving antenna mode in the received signal tensor u; then 1B, 2G, and 3A represent 3 factor matrices;

[0078] where Ip is the first-order identity tensor; N is the corresponding noise tensor; the symbol Xn represents the multiplication of the n-th mode of the tensor and the matrix;

[0079] In step S26, the input tensor is The input relative error is ε, and the output satisfies and H, S, and C are non-negative matrices. The non-negative matrices H, S, and C are randomly initialized; fixing S and C, the update of H is represented as

[0080]

[0081] Fixing H and C, the update of S is represented as

[0082] Fix H and S, and update C as Repeat steps (1), (2), and (3) in a loop until the relative error is less than ε, then stop the iteration.

[0083] In this embodiment, step S3 includes the following steps:

[0084] In step S31, the transmitter r a (k) and r b (k) are a pair of Gray sequences Ga N and Gb N of length N for transmission respectively. Through the multipath channel and additive white Gaussian noise, the received sequences are respectively: where h(k) represents the actual multipath channel impulse response, and "*" represents the convolution operation. n a (k) and n b (k) represent Gaussian noise with a mean of 0 and a variance of σ 2 ;

[0085] In this embodiment, step S3 further includes the following steps:

[0086] In step S32, the received r a (k) and r b (k) are respectively cross-correlated with the local Gray sequences Ga N and Gb N , and the results after the cross-correlation operations are added to obtain the channel impulse response r CIR (k), where where Ra N (k) and Rb N (k) respectively represent the autocorrelation results of the Gray sequences Ga N and Gb N . n(k) represents Gaussian noise with a mean of 0 and a variance of 2Nσ 2 ; At this time, the length of the channel impulse response r CIR (k) is twice the length of the Gray sequence Ga N .

[0087] In this embodiment, step S3 further includes the following steps:

[0088] In step S33, calculate the channel power of the channel impulse response r CIR (k), calculate the average power of the first half of the channel impulse response r CIR (k). This average power is the noise power. Divide the calculated signal power by the estimated noise power to obtain the signal-to-noise ratio, and amplify the channel power by 4N 2 times. At this time, the corresponding increase multiple of the signal-to-noise ratio is 2N.

[0089] Specifically, a time-varying channel model for a millimeter-wave massive MIMO system in a high-speed moving scenario is first established. Due to the large gap between the number of radio frequency (RF) links and the number of antennas in the millimeter-wave hybrid structure, only a low-dimensional projection of the received signal can be observed at the receiving end. A switching network based on analog design is used to select antennas, and a nested sampling strategy is adopted to form a virtual array with a large aperture. This method can significantly reduce the number of RF links and the system computational complexity.

[0090] To achieve real-time channel detection and estimation, the property of complementary Gray sequence pairs is utilized to complete channel estimation in the time domain. Let r a (k) and r b (k) be the results of the N-point Gray sequences Ga N and Gb N after passing through the channel, respectively. According to the property of complementary Gray sequence pairs, as shown in the formula Ra N (k)+Rb N (k) = Nδ(k). Therefore, the multipath channel h(k) can be accurately estimated, and the estimated channel power is amplified by 4N 2 times. Therefore, Gray correlation operation can not only estimate the channel impulse response, but also improve the signal-to-noise ratio (SNR) during channel estimation. The corresponding increase in SNR is 2N.

[0091] For the first-stage tensor decomposition, small tensors are randomly derived from the large tensor, and the high-dimensional channel is projected into the null space of the interfering user channels; for the second-stage third-order tensor decomposition, the traditional parafac tensor factorization method requires a large amount of time and memory consumption when dealing with large-scale problems. Therefore, a grid tensor decomposition method will be used to process large-scale tensors.

[0092] After performing correlation processing on different multipath signals at the receiving end, the interference between multipaths in the received signal is reduced, thereby enhancing its energy and the signal characteristics. Subsequently, a non-negative decision output is performed on the signal, and a random grid tensor decomposition model is formed at the receiving end. Since tensor decomposition emphasizes the non-negativity of the decomposition factors, the data is highly interpretable, and the decomposition result can well represent the local characteristics of the channel state information and is unique. Therefore, it is widely applied to image and multi-dimensional speech processing. And due to the constraint of non-negative grid factors, the error of the signal after each iteration is reduced to a certain extent, thereby accelerating the system to reach a stable state. To address the problem of large computational complexity in channel estimation for multi-user millimeter-wave systems, a random grid tensor decomposition algorithm is used to estimate channel parameters by utilizing the sparse scattering characteristics of millimeter-wave channels and the spatial structure of tensors.

[0093] This method represents the received signal as a fourth-order tensor, transforms the channel parameter estimation problem into a large-scale tensor decomposition problem, uses a random tensor decomposition method for tensor compression, and then uses a grid tensor decomposition algorithm for parallel tensor calculation, reducing the number of inverse and multiplication operations of high-dimensional matrices. This algorithm can obtain accurate channel parameter estimation and effectively reduce the complexity of the channel estimation algorithm.

[0094] In this application, by using the random grid tensor decomposition method, the coherent structure is learned from the large-scale tensor, and the user channel is projected into the orthogonal space of interfering users to suppress the interference between different channels.

[0095] Meanwhile, the compressed user channel tensor is decomposed by using the grid tensor algorithm, and the large-scale tensor is transformed into several small grids to obtain relatively accurate estimated values.

[0096] The simulation results of traditional pilots, parallel factor decomposition, and the proposed method are used to verify the superiority of the proposed method.

[0097] Based on the wideband geometric channel model, a millimeter-wave channel is generated. Among them, when the number of transmitting antennas Nt is 64 and the number of receiving antennas Nr is 64, assuming there are 5 resolvable paths, the time delays τ of the 5 paths are taken as 0.1, 1.0, 1.5, 2.0, 3.0 Ts, where Ts is the symbol period, and the AOA and AOD of the channel are uniformly randomly generated within [0, 2π]. 1000 Monte Carlo simulations are carried out for the experiment.

[0098] First, the comparison of the mean square error (MSE) of the estimation of each algorithm.

[0099] From Figure 4 it can be seen that the proposed method is superior to the traditional pilot method and the parallel factor decomposition method. Especially in the case of high signal-to-noise ratio (SNR), the estimation performance advantage of the proposed method is more obvious.

[0100] Meanwhile, in the proposed method, the performance when the number of sub-tensors M is 4 is slightly better than when M is 2. Therefore, the performance of the random grid algorithm can be improved by appropriately setting parameters.

[0101] Second, the bit error rate (BER) of each algorithm is compared under different signal-to-noise ratios.

[0102] From Figure 5 it can be seen that the bit error rate of this algorithm is lower than that of the other two algorithms. This is because there is a spatial structure of tensors in the random grid algorithm, and the classification of sub-tensors reduces the accumulation of layer-by-layer iteration errors, reduces the hierarchical sub-bit error rate, and improves the overall bit error rate performance of the system.

[0103] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should understand that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0104] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation manner. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present invention.

[0105] Therefore, as described above, the above is only the preferred specific implementation manner of the present invention, and the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. The protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for channel estimation of a millimeter wave MIMO system based on random tensor network decomposition, wherein the transmitting end of the millimeter wave MIMO system is equipped with Nt transmitting antennas, the receiving end is equipped with Nr receiving antennas, the millimeter wave MIMO system has a total of k subcarriers, wherein the k subcarriers are used for channel estimation, characterized in that: The method comprises the following steps: S1, obtain the transmission signal on k subcarriers and the tth time frame, the transmitting antenna sends the transmission signal to the receiving antenna, and the transmission signal is represented by x k (t) = s k (t)c k (t), Among them, S k (t) is represented by the orthogonal pilot signal on the kth subcarrier, C k (t) is represented as the precoding matrix on the kth subcarrier, and the k subcarriers are represented as x(t)=[x1(t),...,x K (t)] T ; S2, using the sparse scattering characteristics of the millimeter wave channel and the spatial structure of the tensor, the receiving antenna obtains the received signal, stacks the received signal into a three-dimensional tensor form, and constructs three factor matrices corresponding to the three-dimensional tensor form. The received signal includes the number of users, receiving antennas, subcarriers and pilot symbols. A random grid tensor decomposition algorithm is used to estimate the channels between the receiving antenna and the subcarrier, the receiving antenna and the pilot symbol, and the receiving antenna and the user to obtain the channel parameters of the first three time slots of the time-varying channel; S3, using the GOLAY sequence Ga N and Gb N The factor matrix is ​​estimated in the time domain respectively to obtain the estimated channel result.

2. The method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition according to claim 1, characterized in that: The step S1 comprises the following steps: S11, P multipaths are set between the transmitting end and the receiving end, and the delay caused by each multipath is τ p , then the space-time channel is expressed as Among them, α p Expressed as the attenuation coefficient on the Pth path, φ p ∈[0,2π] and φ p ∈[0,2π] represent the arrival angle and departure angle on the Pth path, δ(·) represents the impulse response function, τ p represents the delay on the Pth path, α r (·) and a t (·) denote the uniform antenna array responses at the receiving and transmitting ends, respectively.

3. The method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition according to claim 2, characterized in that: The step S1 further comprises the following steps: S12, the receiving end preprocesses the received signal obtained by the receiving antenna, and uses a correlator to receive each multipath received signal. The receiving end preprocesses the P-th multipath received signal of the j-th receiving antenna, and outputs the preprocessed multipath received signal as Z through a matched filter. j，p (t) = ∫r j (t)w(t-τ j，p )dt, where τ j，p It is represented as the correlation delay of the pth multipath signal of the jth receiving antenna, w(t-τ) is represented as the synchronous correlation function of the corresponding multipath delay; Rj(t) represents the jth received signal.

4. The method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition according to claim 3, characterized in that: The step S1 further comprises the following steps: S13, the transmitting end preprocesses the transmission signal of the transmitting antenna to obtain a transmission signal Z j [n], for the transmitted signal Z j [n] Make a decision to get the final output transmission signal Y j , for the transmitted signal Y j Pre-programmed in matrix form in, It is expressed as a matrix expansion along the latitude of the receiving antenna, It is represented as the transmission signal matrix formed by the Kth subcarrier on the Ntth transmitting antenna, SD J (H) represents the diagonalization operation, taking out the j-th row vector H of the channel transmission matrix H J , with H J The elements of are taken as the main diagonal elements, and the other elements are recorded as zero. is represented as a noise matrix whose elements are independent and identically distributed additive Gaussian white noise, It represents the precoding matrix for the transmitted signal, and R represents a non-negative real matrix operation.

5. The method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition according to claim 4, characterized in that: The step S2 comprises the following steps: S21, stacking the received signals of Nr receiving antennas into a three-dimensional tensor form Y, where ⊙ denotes Khatri-Rao product, H denotes row vector, SD J (H) represents the diagonalization operation.

6. The method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition according to claim 5, characterized in that: The step S2 further comprises the following steps: S22, build The corresponding three factor matrices are in, They are represented as column vectors of factor matrices H, S, and C respectively, and Ν represents the corresponding noise tensor.

7. The method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition according to claim 6, characterized in that: The step S2 further comprises the following steps: S23, extract random vectors ω1, ω2, ω3 through Gaussian distribution as the approximate basis of the n-order tensor model, and construct the measurement matrix with random vectors The column space representation for the user channel tensor slice Ωn is S n =y (n) Ω n , n = 1, 2, ..., N, where y(n) is the expansion of the tensor y modulo -n; It is represented as a channel sample matrix, and the sample matrix is ​​used as an approximate basis for the modulo-n expansion tensor; S24, sampling the received signal modulo-n expanded tensor y(n) by random projection method, and decomposing the channel sample matrix Sn into: S n =Q n R n· ; Use Schmidt orthogonal method to construct equivalent channel sample matrix: And through random tensor decomposition, the received signal of each user is obtained, which is expressed as a third-order tensor: Among them, E i Denote as the error objective function under Euclidean distance, C is represented as a set of complex numbers, where each module of the tensor represents the receiving antenna (Nr), subcarrier (K), and symbol (T). S25, assuming that the transmitter sends the same pilot symbol, let The expansion form of a receiving antenna module in the received signal tensor u is: Among them, B, G, A are three factor matrices; N is the noise matrix; is the attenuation coefficient matrix of all Nr receiving antennas; @ is represented by KhatriRao product, and we get: u = I p ×1B×2G×3A+N, let A be a receiving antenna module in the received signal tensor u; then 1B, 2G, 3A are represented as 3 factor matrices; Where Ip is the first-order unit tensor; N is the corresponding noise tensor; the symbol Xn represents the multiplication of the tensor modulo-n and the matrix; S26, the input tensor is The input relative error is ε, and the output satisfies And H, S, C are non-negative matrices, randomly initialize non-negative matrices H, S, C; fix S, C, and update H as Fix H,C and update S as Fix H, S and update C as Repeat steps (1), (2), and (3) until the relative error is less than ε.

8. The method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition according to claim 1, characterized in that: The step S3 comprises the following steps: S31, the transmitting end r a (k) and r b (k) are a pair of Gray sequences Ga sent with a length of N points. N , Gb N , through the multipath channel and Gaussian white noise, the received sequences are: Where h(k) represents the actual multipath channel impulse response, "*" represents the convolution operation, and n a (k) and n b (k) represents a mean of 0 and a variance of σ 2 Gaussian noise.

9. The method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition according to claim 8, characterized in that: The step S3 further comprises the following steps: S32, the received r a (k) and r b (k) respectively and the local Gray sequence Ga N and Gb N Perform cross-correlation operation and add the results of cross-correlation operation to obtain the channel impulse response r CIR (k), where Among them, Ra N (k) and Rb N (k) are respectively represented by Ga N Gray sequence and Gb N The result of Gray sequence autocorrelation, n(k) represents the mean of 0 and the variance of 2Nσ 2 Gaussian noise; at this time, the channel impulse response r CIR (k) is the length of the Gray sequence Ga N Twice the length.

10. The method for realizing millimeter wave MIMO system channel estimation based on random tensor network decomposition according to claim 9, characterized in that: The step S3 further comprises the following steps: S33, calculate the channel impulse response r CIR (k) channel power, calculate the channel impulse response r CIR The average power of the first half of (k) is the noise power. The calculated signal power is divided by the estimated noise power to obtain the signal-to-noise ratio, and the channel power is amplified by 4N. 2 times, at this time, the corresponding signal-to-noise ratio increases by a factor of 2N.

Images