A limited feedback based channel estimation method for MIMO-OFDM systems

Abstract

The embodiment of the present application provides a channel estimation method based on limited feedback for a MIMO-OFDM system, which is applied to a base station end and used for estimating a downlink channel state, and the method comprises the following steps: S1, obtaining a preset vector quantization codebook for the MIMO system, which comprises a plurality of code vectors for representing different pilot received signal vectors and an index value of each code vector; S2, sending a pilot signal to a plurality of user ends and receiving a signal fed back by each user end to the base station end according to a pilot received signal of the user end, wherein the fed back signal is an index value of a code vector most similar to a quantization vector of the pilot received signal of the user end searched by the user end from the vector quantization codebook; S3, obtaining an estimated quantization vector of the pilot received signal according to the vector quantization codebook and the signal fed back by each user end; and S4, performing channel estimation according to the estimated quantization vector by using a preset estimation mode to obtain a downlink channel state information matrix corresponding to each user end.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wireless communication, in particular, to the field of channel estimation, more particularly, to a channel estimation method based on limited feedback for MIMO-OFDM system. BACKGROUND

[0002] Multi-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) technology is widely used in modern wireless communication systems due to its outstanding multiplexing, high data transmission rate and high spectrum utilization. MIMO technology transmits multiple data streams by using multiple transmit and receive antennas in the same frequency band, significantly improving data transmission rate and system capacity. The combination of MIMO technology and OFDM technology can effectively combat frequency-selective fading and further enhance transmission performance; it can also greatly improve spectrum utilization through spatial multiplexing without increasing additional spectrum resources, which is particularly important for today's spectrum-starved wireless communication environment.

[0003] According to whether a pilot signal is needed, the current channel estimation algorithm can be divided into two categories: blind channel estimation and pilot-based channel estimation. Blind channel estimation does not require a pilot signal, and the pilot signal is estimated by using the second-order or high-order statistical information of the received signal. For example, reference [1] proposes a blind channel estimation algorithm that uses the statistical information of the average power of the received signal to convert the average power of the received signal into a quadratic equation containing channel gain. The user terminal can estimate the MIMO downlink channel gain without any downlink pilot resource, but this algorithm is only suitable for time division duplex systems and has poor estimation accuracy. For example, reference [2] proposes a MIMO channel blind estimation algorithm based on expectation maximization, which uses the sparse characteristics of the channel in the angle domain to improve the channel estimation accuracy, but requires a large amount of calculation. Blind channel estimation has the advantages of less prior information and high spectrum efficiency, but its application is severely limited by low channel estimation accuracy, high complexity and poor real-time performance.

[0004] Although pilot-based channel estimation loses some spectral efficiency compared to blind channel estimation, it is widely used due to its simplicity. Commonly used pilot-based channel estimation algorithms include the least square (LS) method in reference [3] and the minimal mean square error (MMSE) in reference [4]. LS algorithm is a simple interpolation-based method, which has poor performance due to the neglect of noise processing. MMSE algorithm considers the complete channel statistics and noise variance, and has better estimation accuracy than LS algorithm, but it needs to invert the channel correlation matrix, which has large computation and needs prior channel statistics. Reference [5] proposes an approximate linear MMSE for fast fading channels, which greatly reduces the complexity of correlation and filter matrix size.

[0005] The basic idea of pilot-based channel estimation technology in MIMO-OFDM systems is as follows: first, the sending end periodically sends a pre-designed pilot signal sequence; then, the receiving end receives the pilot signal and generates a feedback data according to the received signal quality and conditions; finally, the sending end receives the feedback from the receiving end and uses the information to perform more accurate channel estimation. Since MIMO-OFDM systems usually include multiple subcarrier channels and multiple transceiver antenna pairs, the system needs to estimate a large number of channel parameters, and the traditional pilot-based and feedback-based channel estimation algorithm will cause the pilot and feedback overhead to increase sharply with the increase of subcarriers and transceiver antenna pairs, and the performance loss in spectral efficiency and complexity is serious, which limits its application. At present, for MIMO-OFDM systems, scholars have proposed vector quantization-based channel estimation algorithms, such as feeding back quantized channel information or quantized received pilot information, which can reduce the feedback overhead to some extent compared to directly feeding back channel or pilot information, but vector quantization introduces some quantization errors, which further causes the loss of channel estimation performance, and the quantized data information is still a lot, and the feedback overhead is large.

[0006] Therefore, the existing pilot-based channel estimation technology in MIMO-OFDM systems has the problems of large feedback information and large feedback overhead when feeding back channel information, which reduces the efficiency of wireless communication.

[0007] It should be noted that the background art is only used to introduce the relevant information of the present application, so as to help understand the technical solutions of the present application, but does not mean that the relevant information must be prior art. In the absence of evidence that the relevant information has been disclosed before the filing date of the present application, the relevant information should not be considered as prior art.

[0008] Reference:

[0009] [1] Ngo H Q, Larsson E G. Blind estimation of effective downlink channel gains in massive MIMO [C] / / 2015IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2015: 2919-2923.

[0010] [2] Mezghani A, Swindlehurst A L. Blind estimation of sparse broadband massive MIMO channels with ideal and one-bit ADCs [J]. IEEE Transactions on Signal Processing, 2018, 66(11): 2972-2983.

[0011] [3] Chang M X. A new derivation of least-squares-fitting principle for OFDM channel estimation [J]. IEEE transactions on wireless communications, 2006, 5(4): 726-731.

[0012] [4] Minn H, Bhargava V K, Letaief K B. A combined timing and frequency synchronization and channel estimation for OFDM [C] / / 2004IEEE International Conference on Communications (IEEE Cat. No. 04CH37577). IEEE, 2004, 2: 872-876.

[0013] [5] M, Mehlführer C, Wrulich M, et al. Doubly dispersive channel estimation with scalable complexity [C] / / 2010 International ITG Workshop on Smart Antennas (WSA). IEEE, 2010: 251-256. SUMMARY

[0014] Therefore, the purpose of the present application is to overcome the defects of the prior art, to provide a limited feedback-based channel estimation method for MIMO-OFDM system.

[0015] The purpose of the present application is achieved by the following technical solutions:

[0016] According to the first aspect of the present application, a channel estimation method is provided, which is applied at the base station end and used for estimating the downlink channel state, the method comprising: S1, obtaining a preset vector quantization codebook for MIMO system, which comprises a plurality of code vectors for representing different pilot received signal vectors and an index value of each code vector; S2, sending a pilot signal to a plurality of user ends and receiving a signal fed back by each user end to the base station end according to its pilot received signal, the fed back signal being an index value of a code vector searched by the user end from the vector quantization codebook and most similar to the quantization vector of the pilot received signal; S3, obtaining an estimated quantization vector of the pilot received signal according to the vector quantization codebook and the signal fed back by each user end; S4, performing channel estimation according to the estimated quantization vector by using a preset estimation method, to obtain a corresponding estimated downlink channel state information matrix of each user end.

[0017] In some embodiments of the present application, in the S1, the generation manner of the vector quantization codebook comprises: constructing a vector source training set for an indoor communication scenario, including obtaining a plurality of positions in the indoor communication scenario, calculating pilot received signal vectors of users located at the plurality of positions, to obtain the vector source training set; generating a vector quantization codebook by using a vector quantization algorithm according to the vector source training set and a preset quantization bit number, the vector quantization codebook comprising a plurality of code vectors and an index value corresponding to each code vector, the preset quantization bit number being a bit number of the index value and being less than a bit number of the code vector.

[0018] In some embodiments of the present application, the manner of constructing the vector source training set for the indoor communication scenario comprises: generating a plurality of positions of random walking of users in a predetermined indoor environment by using a random walking model; calculating pilot received signal vectors of a plurality of users respectively walking to each position according to a pilot signal transmission and a channel model, to obtain the vector source training set.

[0019] In some embodiments of the present application, in the S4, the preset estimation method comprises: constructing an optimal post-processing matrix optimization function based on a minimum mean square error criterion, with minimizing the gap between the real downlink channel state information matrix and the estimated downlink channel state information matrix as the optimization objective; solving the optimal post-processing matrix based on the optimization function and the estimated quantized vector of the pilot received signal; and obtaining the estimated downlink channel state information matrix according to the optimal post-processing matrix and the estimated quantized vector of the pilot received signal.

[0020] In some embodiments of the present application, the optimal post-processing matrix optimization function is as follows:

[0021]

[0022] wherein, represents the gap between the real downlink channel state information and the estimated downlink channel state information matrix, D represents the post-processing matrix, E{·} represents the expected calculation symbol, ‖·‖ 2 represents the two-norm of the vector, s = vec(H down ), vec(·) represents the matrix vectorization operation, H down represents the real downlink channel state information matrix, represents the estimated downlink channel state information matrix, represents the pilot received signal y pil ot the estimated value of s when known.

[0023] In some embodiments of the present application, the vector quantization algorithm comprises multiple rounds of iteration according to a preset quantization bit number, and each round of iteration comprises: obtaining a code vector, wherein the first round adopts a code vector calculated according to each sample in the vector source training set and the sample quantity, and each round after the first round adopts the code vector obtained in the last round; obtaining a scrambling coefficient, splitting the obtained code vector according to the scrambling coefficient to obtain multiple split code vectors; and updating each code vector in the multiple split code vectors once or multiple times according to a preset rule to obtain the code vector of the current round.

[0024] In some embodiments of the present application, the preset rule comprises: obtaining multiple code vectors at the current time, and the first time adopts the multiple split code vectors, and each time after the first time adopts the multiple code vectors updated in the last time; dividing the vector source training set into multiple sub-sets based on the nearest neighbor rule according to the multiple code vectors at the current time, and each code vector at the current time corresponds to a sub-set; calculating the total distortion degree according to the multiple divided sub-sets, judging whether the total distortion degree converges, and if not, updating the multiple code vectors at the current time to obtain the multiple updated code vectors, including averaging the sum of all samples in the sub-set corresponding to each code vector to obtain the updated code vector, and otherwise stopping the above updating process to obtain the code vector of the current round.

[0025] In some embodiments of the present application, in the S2, the feedback signal is obtained by the user terminal in the following manner: vector quantizing the pilot received signal to obtain a quantized vector of the pilot received signal; searching the vector quantization codebook to find a code vector most similar to the quantized vector of the pilot received signal, and taking the index value of the most similar code vector as the feedback signal.

[0026] In some embodiments of the present application, the pilot received signal is vector quantized in the following manner:

[0027]

[0028] wherein y pilot represents the quantized vector of the pilot received signal, vec(·) represents a matrix vectorization operation, Y pilot represents the pilot received signal, and Φ represents the transmitted pilot signal. represents a tensor product, I Nr represents an N r ×N r identity matrix, N r represents the number of receiving antennas of all user terminals, H down represents the real downlink channel state information matrix, and W pilot represents a noise matrix when the pilot signal is received.

[0029] According to a second aspect of the present application, a communication system based on the method of the first aspect of the present application is provided, the system comprising a plurality of user terminals and a base station terminal, wherein the base station terminal is configured to: obtain a vector quantization codebook preset for the MIMO system, the codebook comprising a plurality of code vectors for representing different pilot received signal vectors and an index value of each code vector; transmit a pilot signal to the plurality of user terminals and receive a signal feedback from each user terminal according to its pilot received signal; obtain an estimated quantized vector of the pilot received signal based on the feedback signal of each user terminal and the vector quantization codebook; perform channel estimation according to the estimated quantized vector using a preset estimation method to obtain an estimated downlink channel state information matrix corresponding to each user terminal; and optimize the communication parameter configuration of the base station terminal according to the estimated downlink channel state information matrix of each user terminal, and transmit a data signal to each user terminal based on the optimized communication parameter configuration; and each user terminal is configured to: obtain the vector quantization codebook preset for the MIMO system, search the vector quantization codebook to find an index value of a code vector most similar to the quantized vector of its pilot received signal according to its pilot received signal, and feed back the index value of the most similar code vector to the base station terminal as the signal; and receive the data signal transmitted by the base station terminal and obtain data according to its own position and the received data signal.

[0030] Compared with the prior art, the present application has the following advantages:

[0031] The present application sends pilot signals to multiple user terminals through a base station end, vector quantizes pilot receiving signals of all user terminals, and feeds back an index value of a code vector most similar to a searched quantization vector from a vector quantization codebook to the base station end through an uplink, avoids feedback of channel information or pilot signals and other parameter information, and realizes estimation only by feeding back an index value, greatly reducing feedback data and feedback cost of the present application. Finally, the base station end of the present application can directly obtain an estimated quantization vector of a pilot receiving signal from a vector quantization codebook according to a signal fed back by each user terminal, and estimates a downlink channel state information matrix corresponding to each user terminal according to the estimated quantization vector, reduces calculation complexity, and improves performance and efficiency of wireless communication. BRIEF DESCRIPTION OF DRAWINGS

[0032] The embodiments of the present application are further described below with reference to the accompanying drawings, in which:

[0033] Figure 1 It is a schematic diagram of a transmitting and receiving process principle of a MIMO-OFDM system model according to an embodiment of the present application;

[0034] Figure 2 It is a schematic diagram of a channel estimation method flow based on limited feedback according to an embodiment of the present application;

[0035] Figure 3 It is a schematic diagram of a block pilot transmitting signal structure adopted according to an embodiment of the present application;

[0036] Figure 4 It is a schematic diagram of a pilot transmission and feedback flow in a MIMO-OFDM system according to an embodiment of the present application;

[0037] Figure 5 It is a flow chart of a channel estimation method based on limited feedback for a MIMO-OFDM system according to an embodiment of the present application;

[0038] Figure 6 It is a comparison diagram of mean square error curves corresponding to channel state information matrices estimated by six schemes varying with signal noise ratio SNR according to an embodiment of the present application;

[0039] Figure 7 It is a comparison diagram of mean square error curves varying with code vector number generated by the method of the present application and a traditional vector quantization scheme when estimating channels under different signal noise ratios SNR according to an embodiment of the present application;

[0040] Figure 8 It is a comparison diagram of calculation time delays of different numbers of code vectors generated by the method of the present application and a traditional vector quantization scheme and a scalar quantization scheme according to an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0042] As mentioned in the section of background art, the existing pilot-based channel estimation techniques for MIMO-OFDM systems have the problems of too much feedback information and too much feedback overhead when feeding back channel information, which reduces the efficiency of wireless communication.

[0043] Based on the above problems, according to one embodiment of the present application, the inventors propose a pilot-based channel estimation method for MIMO-OFDM systems, which is used to estimate the downlink channel state. In the method of the present application, first, a vector quantization codebook is preset for the MIMO system, which includes a plurality of code vectors for representing different pilot reception signal vectors and an index value of each code vector; then, the base station end of the present application sends pilot signals to a plurality of user ends, vector quantizes the pilot reception signals of all user ends, and feeds back the index value of the code vector most similar to the quantized vector searched from the vector quantization codebook to the base station end through the uplink, so as to avoid feeding back a plurality of parameter information such as channel information or pilot signals, and only feed back an index value to realize estimation, which greatly reduces the feedback data and feedback overhead of the present application. Finally, the base station end of the present application can directly obtain the estimated quantized vector of the pilot reception signal from the vector quantization codebook according to the signal fed back by each user end, and estimate the downlink channel state information matrix corresponding to each user end according to the estimated quantized vector, which reduces the calculation complexity and improves the performance and efficiency of wireless communication.

[0044] Before describing the channel estimation method of the present application, the MIMO-OFDM system model is introduced.

[0045] According to one embodiment of the present application, a MIMO-OFDM system model is set, in which the base station end uses N t transmit antennas, and the user reception end uses N r receive antennas. Generally, one user end is one receive antenna, that is, N r user ends form a user reception end to receive data. Therefore, N t ×N r independent parallel space transmission channels are formed between the base station end and all user ends, and the MIMO-OFDM system uses OFDM technology and the total number of OFDM subcarriers is K. Referring to Figure 1 , which is a schematic diagram of the transmission and reception process principle of the MIMO-OFDM system model, the transmission and reception process principle of the MIMO-OFDM system model will be described below with reference to Figure 1 .

[0046] (1) Transmission process principle of MIMO-OFDM system model

[0047] According to one embodiment of the present invention, Figure 1 As shown, the sending end is the base station end, the 1st to the Nth t The signals sent by the transmitting antennas are recorded as For the nth t Transmitting antennas, n t =1,…,N t , in the nth OFDM symbol interval, the constellation modulation signal obtained by the transmitter after serial-to-parallel conversion can be expressed as k represents the number of OFDM subcarrier, Indicates the nth t The constellation modulation signal of the kth OFDM subcarrier of the nth OFDM symbol on the transmitting antenna, K represents the number of OFDM subcarriers, and the nth OFDM symbol can be obtained by inverse fast Fourier transform (IFFT). t The OFDM time domain signal on the transmitting antenna can be expressed as follows:

[0048]

[0049] in, Indicates the nth t The nth OFDM time domain signal on the transmitting antenna, IFFT{·} represents the inverse fast Fourier transform operation, N represents the number of OFDM symbols in the frequency domain. To prevent inter-symbol interference (ISI) in wireless channels, a cyclic prefix (CP) is typically added to each OFDM time-domain signal. This invention considers adding an appropriate guard interval to the transmitted signal in a communication system, while ignoring ISI and synchronization errors (primarily referring to synchronization errors during data signal reception and processing). The transmitter then performs parallel-to-serial conversion on the OFDM time-domain signal with the CP added before transmitting it.

[0050] (2) Principle of the receiving process of the MIMO-OFDM system model

[0051] Assume that the wireless channel is a static or quasi-static channel and remains unchanged in each OFDM symbol period. Then, in the nth OFDM symbol period, after removing the cyclic prefix (CP), the nth OFDM symbol at the user receiving end is r The time domain baseband equivalent signal received by the receiving antenna It can be expressed as follows:

[0052]

[0053] in, Indicates the nth t transmit antennas and the nth r The channel gain between the root receiving antennas, Indicates the nth OFDM symbol period r The Gaussian white noise in the signal received by the root receiving antenna. r After the time domain baseband equivalent signal after CP removal received by the transmitting antenna is subjected to serial-to-parallel conversion, fast Fourier transform (FFT) and parallel-to-serial conversion, the received signals are recorded as Among them, N at the user receiving end r Baseband equivalent time domain signal matrix received by the root receiving antenna It can be expressed as follows:

[0054]

[0055] Where P represents the transmission power of the signal, N represents the downlink in the MIMO-OFDM system model r ×N t The real downlink channel state information matrix of dimension is Indicates that N is included r ×N t The set of real numbers with elements, represents K×N t dimensional transmission signal matrix, W represents the noise matrix when receiving the signal, which is composed of N r ×K are independent of each other and have a mean of 0 and a variance of The variable composition of (·) T The real downlink channel state information matrix can be understood as: the downlink channel state information matrix set by the simulation, which can be calculated based on the relative spatial position and posture of the receiving antenna and the transmitting antenna.

[0056] According to one embodiment of the present invention, see Figure 2 , which is a flow chart of the limited feedback-based channel estimation method of the present invention. The channel estimation method is applied at the base station to estimate the downlink channel state. The estimation method includes steps S1, S2, S3, and S4. To better understand the present invention, each step is described in detail below in conjunction with specific embodiments.

[0057] In step S1, a vector quantization codebook preset for a MIMO system is obtained, which includes a plurality of code vectors for representing different pilot received signal vectors and an index value of each code vector.

[0058] According to one embodiment of the present application, the pilot transmission and reception process of the MIMO-OFDM system model is shown as follows:

[0059] Transmission process: refer to Figure 3 , which is a schematic diagram of the block pilot transmission signal structure adopted, showing the block pilot transmission signal structure corresponding to the channel with the characteristics of time direction and frequency direction. The present application inserts block pilots in the time direction at intervals, such as the black solid circle part is the inserted block pilot, and the black hollow circle part is the inserted data, and the receiving end obtains the pilot received signal. Then, let τ pilots be placed in each OFDM symbol, and be placed at intervals in time slots n K-1 , the nth t element of the nth row of the transmission signal matrix X

[0060]

[0061] , where, represents the signal transmitted by the nth t transmit antenna in the nth time slot, t represents the pilot signal transmitted by the nth t transmit antenna with a scaling factor of m, k represents the data signal transmitted by the nth k transmit antenna with a scaling factor of m and an offset factor of d, t is the number of sub-slots between adjacent pilots, K represents the total number of subcarriers, and the frequency domain OFDM symbol can obtain K time domain complex signals after K-point IFFT transformation as shown in formula (1). The present application adopts block pilot insertion, and let the scaling factor be represented as m. When n is an integer multiple of D k , the nth k transmit antenna transmits a pilot signal, and the transmitted pilot signal is represented as On the contrary, when n = mD t +d and the offset factor d = 1, …, D 1,1 -1, the nth τ,1 transmit antenna transmits a data signal, and the transmitted data signal is represented as

[0062] Reception process: according to formula (3), the pilot received signal can be represented as follows:

[0063]

[0064] , where P represents the transmission power of the signal, represents the transmitted pilot signal, which is in matrix form φ 1,1 denotes a first pilot signal transmitted by a first transmit antenna, φ τ,1 denotes a τth pilot signal transmitted by the first transmit antenna, denotes a first pilot signal transmitted by an n t th transmit antenna, denotes a τth pilot signal transmitted by the n t th transmit antenna, τ denotes a number of pilots in each OFDM symbol, W pilot denotes a noise matrix when receiving the pilot signals, which is composed of N r × τ variables independent of each other and subject to a mean of 0 and a variance of .

[0065] According to an embodiment of the present application, based on the pilot transmission and reception process principles introduced in the above embodiments, the generation manner of the vector quantization codebook in step S1 of the present application is described. The present application considers an indoor mobile communication scenario, and presets a vector quantization codebook for a MIMO system. Therefore, in step S1, the generation manner of the vector quantization codebook includes steps S11 and S12. In step S11, a vector source training set is constructed for the indoor communication scenario, including obtaining multiple positions in the indoor communication scenario, calculating pilot reception signal vectors of users located at the multiple positions, and obtaining the vector source training set. In step S12, the vector quantization codebook is generated by using a vector quantization algorithm according to the vector source training set and a preset quantization bit number, which includes multiple codevectors and index values corresponding to each codevector, and the preset quantization bit number is the bit number of the index value and is less than the bit number of the codevector. The technical scheme of this embodiment can at least achieve the following beneficial technical effects: the communication of the indoor scenario is simulated to construct the source training data set and the codebook, which is convenient for application in the actual indoor scenario. The quantization bit number refers to the number of bits used to represent the index value of the codevector in the quantization codebook, which determines the bit number of the index value. The bit number of the index value of the present application is much smaller than the bit number of the codevector, effectively reducing the feedback overhead.

[0066] According to an embodiment of the present application, in step S11, the manner of constructing the vector source training set for the indoor communication scenario can be: generating multiple positions of the user randomly walking in a predetermined indoor environment by using a random walk model; calculating pilot reception signal vectors of multiple users walking to each position respectively according to the pilot signal transmission and channel model, and obtaining the vector source training set. Wherein, from the top view, the shape of the predetermined indoor environment is rectangular, square or circular, and the channel model can be a channel model based on Rayleigh fading. Wherein, the vector source training set is denoted as y1, y2, …, y T T respectively, each sample is a pilot reception signal vector.

[0067] According to one embodiment of the present application, in step S11, the way of constructing the vector source training set for the indoor communication scenario can also be: randomly selecting a plurality of positions in the predetermined indoor environment by hand, calculating the pilot receiving signal vectors of a plurality of users respectively walking to each position according to the pilot signal transmission and the channel model, and obtaining the vector source training set.

[0068] According to one embodiment of the present application, the smaller the number of quantization bits is, the smaller the feedback overhead is, but the estimation performance of the channel quantization feedback can be poor, and vice versa. Therefore, the present application presets the number of quantization bits according to actual requirements, for example, the number of quantization bits is 6, 8 or 10, etc., to reduce the feedback overhead as much as possible while ensuring the estimation performance.

[0069] According to one embodiment of the present application, in step S12, under the constraints of the given vector source training set and the preset number of quantization bits B, the vector quantization algorithm includes a plurality of rounds of iteration according to the preset number of quantization bits, and each round of iteration includes steps S121, S122 and S123:

[0070] Step S121: obtaining a code vector, wherein the first round uses the code vector calculated according to the samples in the vector source training set and the number of samples, and each round after the first round uses the code vector obtained in the last round.

[0071] According to one embodiment of the present application, in the first round, let the number of code vectors N = 1, and the code vector obtained in the first round is as follows:

[0072]

[0073] wherein, represents the code vector calculated in the first round, card(·) represents the number of samples in the vector source training set, t represents the number of samples, and y t represents the t-th sample. In the first round, the total distortion degree is calculated based on the calculated code vector, and the total distortion degree calculation method in the first round is as follows:

[0074]

[0075] wherein, τ represents the number of pilots of each OFDM symbol, and ||·|| represents the square of the Euclidean distance. 2

[0076] Step S122: obtaining a scrambling coefficient, splitting the obtained code vector according to the scrambling coefficient to obtain a plurality of split code vectors.

[0077] ​According to one embodiment of the present application, the disturbing factors (1+ε) and (1-ε) are set, and the code vectors are split and calculated, and the split and calculation process is as shown in the following formula (8) and formula (9):

[0078]

[0079] wherein, represents the gth code vector after splitting in the 0th iteration (also referred to as the first iteration), represents the g+Nth code vector after splitting in the 0th iteration. g=1,…,N, and N represents the total number of code vectors obtained in the iteration round, represents the gth code vector obtained in the iteration round. Illustratively, the code vectors in the first iteration are split into According to formula (8), the code vector According to formula (9), the code vector At this time, is split into two code vectors, and N takes the value of 1. In the next round of splitting, the code vectors and are split and calculated respectively to obtain four code vectors, and N takes the value of 2.

[0080] According to one embodiment of the present application, the iteration round is equal to the preset number of quantization bits. If the preset number of quantization bits is B, the iteration round is B times, and finally the codebook includes 2 B code vectors, and the corresponding index values are represented by B bits. Illustratively, if the preset number of quantization bits is 2, the iteration round is 2 times, and finally the codebook includes four code vectors, and the corresponding index values are represented by 2 bits, which are 00, 01, 10 and 11 respectively. If the preset number of quantization bits is 3, the iteration round is 3 times, and finally the codebook includes eight code vectors, and the corresponding index values are represented by 3 bits, which are 000, 001, 010, 011, 100, 101, 110 and 111 respectively.

[0081] Step S123: updating each code vector in the split plurality of code vectors according to a preset rule one or more times to obtain the code vectors in the iteration round.

[0082] According to one embodiment of the present application, the preset rule includes the following steps S1231, S1232 and S1233:

[0083] Step S1231: obtaining the plurality of code vectors in the iteration round, and using the split plurality of code vectors in the first time, and using the plurality of code vectors updated in the previous time each time after the first time.

[0084] According to one embodiment of the present application, the plurality of code vectors obtained in the first time are used in the step S122, and if it is the first iteration round, the code vectors and The next round iteration is the first round, the first time to use the code vector and respectively.

[0085] Step S1232: According to the plurality of code vectors of the time, the vector source training set is divided into a plurality of subsets based on the nearest neighbor rule, and each code vector of the time corresponds to a subset.

[0086] According to an embodiment of the present application, the vector source training set is divided into n subsets based on the nearest neighbor rule , and the plurality of code vectors obtained in step S122 are used for the first time. For each sample y t in the vector source training set , find the minimum value of y in all code vectors of the time, denotes the nth code vector after the i-th iteration of splitting, and the plurality of subsets are divided in this way, so that each sample in the subset is closer to the code vector than to any other code vector. Each subset is represented as follows:

[0087]

[0088] wherein, denotes the nth subset, denotes the Euclidean distance between the sample y t and the code vector is smaller than the distance between the sample and any other code vector, denotes all code vectors of the time except the code vector . Since is the closest to the sample y t , the code vector can be approximated as the sample y t .

[0089] According to an embodiment of the present application, the code vector closest to the sample y t represents the sample y t , which is specifically represented as wherein Q(·) represents a vector quantization function. The present application represents the pilot received signal vector by the code vector, thereby generating a vector quantization codebook, and subsequently searching for the code vector most similar to the quantized vector of the pilot received signal from the codebook, thereby facilitating signal feedback and operation by the index value of the most similar code vector.

[0090] Step S1233: Calculate the total distortion according to the divided multiple sub-sets, judge whether the total distortion converges, if not, update the multiple code vectors of this time, obtain the updated multiple code vectors, including averaging the sum of all samples in the sub-set corresponding to each code vector to obtain the updated code vector, otherwise stop the above updating process, and obtain the code vector of this round.

[0091] According to one embodiment of the present application, the total distortion is calculated according to the multiple sub-sets divided at this time in the following way:

[0092]

[0093] Wherein, the total distortion no longer decreases obviously or the number of updates meets the expected threshold, stop updating the code vector. Otherwise, update the multiple code vectors of this time, and repeat the process of steps S1231 to S1233 to realize multiple updates of the code vector. Since the present application represents the pilot received signal vector by the code vector, the present application updates the code vector by minimizing the distance between the sample in each sub-set and the code vector corresponding to the sub-set, so as to iteratively calculate the code vector which can more closely represent the sample (i.e. the pilot received signal vector).

[0094] According to one embodiment of the present application, the way of updating each code vector is as follows:

[0095]

[0096] Wherein, represents the n-th code vector updated at this time. The multiple code vectors obtained by the last update are taken as the code vectors obtained at this round. All code vectors in the vector quantization codebook are the multiple code vectors obtained by the last iteration.

[0097] In step S2, pilot signals are sent to multiple user terminals, and signals fed back by the multiple user terminals to the base station end according to their pilot received signals are received. The fed back signal is the index value of the code vector most similar to the quantization vector of the pilot received signal searched by the user terminal from the vector quantization codebook.

[0098] According to one embodiment of the present application, in the S2, the fed back signal is obtained by the user terminal in the following way: vector quantization is performed on the pilot received signal to obtain the quantization vector of the pilot received signal; the code vector most similar to the quantization vector of the pilot received signal is searched from the vector quantization codebook, and the index value of the most similar code vector is taken as the fed back signal.

[0099] According to one embodiment of the present application, first, pilot signals based on combined codes are selected, and the number of channels required for using combined codes is τ = C(N t ,N tθ), where θ represents the ratio of the maximum power to the average power of the transmit antennas. Then, based on the downlink communication, a pilot signal Φ is transmitted using N t transmit antennas, passes through the downlink channel to the receiving end, and the pilot reception signal is quantized. In the same way as the above formula (5), the pilot reception signal matrix and can be expressed as The formula (5) is equivalent to a vector expression, i.e., the vector quantization of the pilot reception signal is as follows:

[0100]

[0101] where y pilot represents the quantized vector of the pilot reception signal, vec(·) represents the vectorization operation of the matrix, Y pilot represents the pilot reception signal, Φ represents the transmitted pilot signal, represents the tensor product, represents the N r ×N r identity matrix, N r represents the number of receiving antennas of all user terminals, H down represents the real downlink channel state information matrix, W pilot represents the noise matrix when receiving the pilot signal.

[0102] Based on the above operation, the matrix expression of the pilot signal transmitted to the user receiving end through the downlink can be equivalent to the vector form. Define then the vector form is as follows:

[0103]

[0104] Then, the vector form y pilot of (14) is further scalar quantized and expressed as follows:

[0105]

[0106] where Q B (·) represents the scalar quantization, c n represents the n-th code vector in the vector quantization codebook The corresponding index value of the code vector c n is represented as i n , The quantized vector of the pilot reception signal y pilot is defined as the code vector c pilot with the highest similarity to the quantized vector of the pilot reception signal y n , i.e., the code vector c nThe pilot reception signal y pilot may be approximately expressed as a quantized vector, and the index value i n corresponding to the code vector with the highest similarity to the quantized vector of the pilot reception signal y pilot is fed back to the base station end.

[0107] According to an embodiment of the present application, the manner of searching for the code vector c n with the highest similarity to the quantized vector of the pilot reception signal y pilot includes: calculating the similarity between the quantized vector y pilot and each code vector in the vector quantization codebook, and selecting a code vector with the highest similarity to the quantized vector to represent the quantized vector of the pilot reception signal y pilot .

[0108] In step S3, the estimated quantized vector of the pilot reception signal is obtained according to the vector quantization codebook and the signal fed back by each user end;

[0109] According to an embodiment of the present application, the index value i n of the quantized vector of the pilot reception signal for channel state estimation is fed back to the base station end through the uplink communication link, and the received index value y index is expressed in the following form:

[0110] y index = H up P up i n +N, (16)

[0111] wherein H up represents the uplink channel state information matrix, P up represents the precoding matrix, which is used for linear transformation of the signal at the sending end to eliminate the interference between multiple antennas, and N represents the uplink channel noise. The base station end detects the received y index , and obtains the estimation of the index value, which is expressed in the following form:

[0112]

[0113] wherein represents the estimation of the index value, represents the calculation of the generalized inverse of the matrix.

[0114] Finally, the is demapped to obtain the estimated quantized vector of the pilot reception signal. The demapping can be expressed as: wherein represents the code vector searched from the vector quantization codebook according to the estimated index value , and represents the estimated quantized vector of the pilot reception signal.

[0115] In step S4, a preset estimation manner is adopted to perform channel estimation according to the estimated quantized vector, so as to obtain an estimated downlink channel state information matrix corresponding to each user terminal.

[0116] According to an embodiment of the present application, in the step S4, the preset estimation manner comprises the following steps S41, S42 and S43:

[0117] Step S41: An optimal post-processing matrix optimization function based on a minimum mean square error criterion (MMSE criterion) is constructed, with the optimization target being to minimize the gap between the real downlink channel state information matrix and the estimated downlink channel state information matrix.

[0118] According to an embodiment of the present application, the constructed optimal post-processing matrix optimization function is as follows:

[0119]

[0120] Wherein, represents the gap between the real downlink channel state information matrix and the estimated downlink channel state information matrix, D represents the post-processing matrix, E{·} represents an expected calculation symbol, and ‖·‖ 2 represents the two-norm of a vector, s = vec(H down ), vec(·) represents a matrix vectorization operation, H down represents the real downlink channel state information matrix, represents the estimated downlink channel state information matrix, represents the pilot received signal y pilot , and the estimated value of s is known.

[0121] The process principle of constructing the optimal post-processing matrix optimization function is described as follows:

[0122] 1) Analysis of the process of vector quantization of the pilot received signal

[0123] According to an embodiment of the present application, based on the Bussgang decomposition theorem, the process of vector quantization of the pilot received signal in step S2 can be represented as a linear process, i.e., based on formula (15) in step S2, and further represented as the following linear process:

[0124] c n = Gy pilot + e, (19)

[0125] Wherein, G = E{c n (y pilot ) T} E{y pilot (y pilot )T} -1 is a coefficient matrix, e represents a weight vector, and E{·} represents a mean (expectation) calculation symbol.

[0126] 2) Constructing an optimal post-processing matrix optimization function based on the above linear process

[0127] According to one embodiment of the present application, since the purpose is to design an optimal post-processing matrix D to minimize the gap between the real downlink channel state information matrix and the estimated downlink channel state information matrix. Therefore, the optimal post-processing matrix optimization function is initially represented as And the process of vector quantization of the pilot received signal can be represented as a linear process, so the optimal post-processing matrix optimization function can be represented as which can be further decomposed into the following form:

[0128]

[0129] wherein, represents the pilot received signal y pilot The estimated value of s is known, which is independent of the matrix D.

[0130] Step S42: Based on the optimization function and the estimated quantization vector of the pilot received signal, the optimal post-processing matrix is solved.

[0131] According to one embodiment of the present application, define which can be further represented in the following form:

[0132]

[0133] wherein, represents the N r ×N r order unit matrix. In addition, the expansion expression of the second term on the right side of equation (18) can be represented in the following form:

[0134]

[0135] First-order derivation is performed on equation (22), and the first-order derivative can be represented in the following form:

[0136]

[0137] Let and assume that the error between the estimated quantization vector of the pilot received signal and the real quantization vector c n of the pilot received signal can be ignored, that is, At this time, the optimal post-processing matrix is solved, that is, the closed-form solution of the optimal post-processing matrix is as follows:

[0138]

[0139] wherein D represents the optimal post-processing matrix obtained by solving.

[0140] Step S43: obtaining the estimated downlink channel state information matrix according to the optimal post-processing matrix and the estimated quantized vector of the pilot received signal.

[0141] According to an embodiment of the present application, the estimated downlink channel state information matrix is obtained by and is in the form of

[0142]

[0143] wherein Mat{·} represents an operator symbol for converting a vector into a matrix form.

[0144] The technical solution of the above embodiment of step S4 can achieve at least the following beneficial technical effects: since the existing estimated channel state is and the estimated channel state according to the present application can be further to improve the estimation accuracy. Therefore, the present application adopts the vector quantization and the minimum mean square error estimation method, represents the vector quantization as a linear process based on the Bussgang theorem, and further deduces the closed-form solution of the post-processing matrix designed based on the minimum mean square error criterion (MMSE criterion), so as to achieve the technical effect of low complexity, low overhead and high estimation accuracy of the channel estimation, and has the advantages of reducing the feedback overhead of the channel estimation and improving the accuracy of the channel estimation.

[0145] According to one embodiment of the present invention, a communication system based on the method described in the above embodiment is provided, comprising a plurality of user terminals and base station terminals. The base station is configured to: obtain a vector quantization codebook preset for the MIMO system, which includes multiple code vectors for representing different pilot received signal vectors and an index value of each code vector; send pilot signals to multiple user terminals, and receive signals fed back by multiple user terminals based on their pilot received signals; obtain an estimated quantization vector of the pilot received signal based on the signal fed back by each user terminal and the vector quantization codebook; perform channel estimation based on the estimated quantization vector using a preset estimation method to obtain an estimated downlink channel state information matrix corresponding to each user terminal; and optimize the communication parameter configuration of the base station based on the estimated downlink channel state information matrix corresponding to each user terminal, and send a data signal to each user terminal based on the optimized communication parameter configuration; each user terminal is configured to: obtain a vector quantization codebook preset for the MIMO system, search the vector quantization codebook for the index value of the code vector most similar to the quantization vector of its pilot received signal based on its pilot received signal, and feed back the index value of the most similar code vector as a signal to the base station; and receive the data signal sent by the base station, and obtain data based on its own position and the received data signal.

[0146] According to one embodiment of the present invention, during the communication process between the base station and the user terminal, the present invention first considers the scenario that the base station needs to know the downlink channel state, and then optimize the communication parameter configuration such as power allocation and beamforming based on it. Therefore, the downlink channel state is estimated first. The estimation process is described in detail in the following sections. Figure 4 , which is a schematic diagram of the pilot transmission and feedback process in the MIMO-OFDM system. In the downlink: the base station broadcasts the pilot signal through the antenna to send the pilot signal Φ to each user terminal, and the user terminal receives the pilot signal through the downlink channel H down The pilot signal transmitted is used to obtain the pilot reception signal y p At this time, the user end performs vector quantization on its pilot reception signal, that is, executes the process of the above formulas (13)-(15) to obtain the pilot reception signal y pilot Quantized vector And search the vector quantization codebook for the index value i of the code vector that is most similar to the quantization vector of the pilot received signal n , and transmit the index value to the base station. In the uplink: the user terminal transmits the index value i n After being fed back to the base station as a feedback signal, the base station receives the signal through the uplink channel H. up The transmitted feedback signal is used to obtain the received feedback signal Y q At this time, the base station executes the formula (17) of the above embodiment to calculate Y q Perform feedback signal detection to obtain the estimated index value right demapping to obtain an estimated quantized vector of the pilot received signal The base station end constructs an optimal post-processing matrix optimization function, and solves the optimal post-processing matrix D according to the function and and the optimal post-processing matrix D to obtain the downlink channel state information matrix Then, the base station end optimizes the communication parameter configuration of the base station end according to the estimated downlink channel state information matrix of each user end, and sends data signals to each user end based on the optimized communication parameter configuration. Finally, each user end receives the data signals sent by the base station end, and obtains data according to its own position and the received data signals.

[0147] According to one embodiment of the present application, based on the above embodiment, a complete process of the channel estimation method of the present application is described. Referring to Figure 5 , which is a flowchart of the channel estimation method based on limited feedback for MIMO-OFDM system, including the following steps:

[0148] 1. Construct a MIMO-OFDM system model, and the base station end and the user receiving end use N t transmit antennas and N r receive antennas, respectively;

[0149] 2. The base station end inserts block pilots in the time direction;

[0150] 3. Randomly generate a plurality of positions based on the random walk model, then calculate the pilot received signal vectors of a plurality of users respectively walking to each position, and store them in the vector source training set. Under the constraint of the number of quantization bits, generate a vector quantization codebook based on the vector quantization algorithm;

[0151] 4. The base station end broadcasts pilot signals to the user end through the transmit antennas, and the pilot signals reach the user end through the downlink communication channel. The user end performs vector quantization on the pilot received signal;

[0152] 5. Based on the Bussgang decomposition theorem, the vector quantization process of the pilot received signal is represented as a linear process;

[0153] 6. Use the uplink communication channel to feed back the index value corresponding to the code vector most similar to the quantized vector of the pilot received signal to the base station end;

[0154] 7. Construct an optimal post-processing matrix optimization function based on the MMSE criterion, and solve the estimated downlink channel state information matrix according to the feedback index value and the function.

[0155] To verify the beneficial effects of the present application, the inventors conducted the following simulation comparison experiments: ​

[0156] 1) Different code vectors, the change of the mean square error generated by channel estimation when the signal-to-noise ratio increases

[0157] Parameter setting: In this simulation experiment, the preset quantization bit number B is set to 3 bits, 4 bits and 5 bits respectively, and the code vector number corresponding to the quantization bit number of 3, 4 and 5 is 8, 16 and 32 respectively.

[0158] Scheme setting: Six schemes are constructed for comparison, among which, three schemes are the method of the application constructed under the three different preset quantization bit numbers, and the method of the application is combined with vector quantization of pilot received signals and solving of the optimal post-processing matrix. The other three schemes are the traditional vector quantization scheme constructed under the three different preset quantization bit numbers, and the traditional vector quantization scheme only performs vector quantization on pilot received signals.

[0159] Comparison result: The above six schemes are simulated and compared, and the comparison result is shown in Figure 6 , which is a comparison diagram of the mean square error of the downlink channel state information matrix estimated by the six schemes corresponding to the change of the signal-to-noise ratio SNR. The abscissa is the signal-to-noise ratio, and the ordinate is the mean square error. According to Figure 6 , in this simulation experiment, the channel estimation mean square error performance of the traditional vector quantization scheme and the method of the application is improved with the increase of the code vector number, which shows that the more the quantization bit number is, the more accurate the quantization is. The change trend is consistent with the theoretical analysis. By vertically comparing the two algorithms, it can be found that under the same signal-to-noise ratio and code vector number, the mean square error between the downlink channel state information estimated by the method of the application combined with vector quantization and optimal post-processing matrix and the real downlink channel state information is smaller than that of the traditional vector quantization scheme, which proves that the method of the application can greatly improve the quantization performance and reduce the error caused by the traditional vector quantization. At the same time, the quantization effect can be improved by appropriately increasing the code vector number.

[0160] 2) Different signal-to-noise ratios, the change of the mean square error generated by channel estimation when the code vector number increases

[0161] Scheme setting: Six schemes are constructed for comparison, among which, three schemes are the method of the application constructed under the three different signal-to-noise ratios, and the method of the application is combined with vector quantization of pilot received signals and solving of the optimal post-processing matrix. The other three schemes are the traditional vector quantization scheme constructed under the three different signal-to-noise ratios, and the traditional vector quantization scheme only performs vector quantization on pilot received signals.

[0162] Comparison result: The above six schemes are simulated and compared, and the comparison result is shown inFigure 7 As shown in the figure, which is a comparison diagram of the mean square error generated by the method of the application and the traditional vector quantization scheme when estimating the channel under different signal-to-noise ratios (SNRs), the abscissa is the code vector number, and the ordinate is the mean square error. According to the formula Figure 7 It can be seen that when the traditional vector quantization scheme estimates the channel, the estimation effect does not significantly improve with the increase of the signal-to-noise ratio, and the mean square error only slowly decreases; on the contrary, each increase of the code vector number (increase of the quantization bit number) will make the quantization effect improve by one level. However, in the method of the application, when the quantization bit number reaches 4 bits or more, the increase of the signal-to-noise ratio will significantly optimize the estimation effect, and the increase of the quantization bit number will also make the optimization effect improve more greatly, which further indicates that the method of the application has great estimation superiority under both low and high signal-to-noise ratios. In the signal-to-noise ratio range of 0 to 30 dB, the method of the application has at least 80% optimization of the channel estimation algorithm of the traditional vector quantization scheme in terms of the mean square error, and the optimization is increasing as the signal-to-noise ratio increases.

[0163] 3) Under different signal-to-noise ratios, the change of the calculation time delay generated when estimating the channel with the increase of the code vector number Parameter setting: set the signal-to-noise ratio (SNR) to 10 dB and 30 dB, respectively.

[0164] Scheme setting: construct 6 schemes for comparison, among which, 2 schemes are the method of the application constructed under the above 2 different signal-to-noise ratios, and the method of the application is combined with vector quantization of the pilot received signal and solving of the optimal post-processing matrix. Another 2 schemes are the traditional vector quantization scheme constructed under the above 2 different signal-to-noise ratios, and the traditional vector quantization scheme only performs vector quantization on the pilot received signal. The last 2 schemes are the traditional scalar quantization scheme constructed under the above 2 different signal-to-noise ratios, and the traditional scalar quantization scheme only performs scalar quantization on the pilot received signal.

[0165] Comparison result: the above 6 schemes are simulated and compared, and the simulation experiment is timed for each algorithm implementation part. Under the initial conditions of the signal-to-noise ratio of 10 dB and 30 dB, 100 tests are performed under different quantization bit numbers, and each experiment is averaged for about 120 iterations, and the comparison result is shown in the figure Figure 8 As shown in the figure, which is a comparison diagram of the calculation time delay generated by the method of the application and the traditional vector quantization scheme and the scalar quantization scheme for different numbers of code vectors. The abscissa is the code vector number, and the ordinate is the calculation time delay, according to the formula Figure 8It can be known that two curves obtained by the same method under different signal-to-noise ratios are almost coincident, and it can be seen that the signal-to-noise ratio has no significant effect on the quantization time, therefore, the curves under different signal-to-noise ratios can be analyzed together. The running time of the method is 0.97 seconds, 1.12 seconds, 1.39 seconds, 1.95 seconds and 3.08 seconds respectively when the code vector number is 4, 8, 16, 32 and 64, which is reduced by 53.83% compared with the average time delay of all code vectors in the traditional scalar quantization scheme, that is, the operation rate of the method is improved by about one time compared with the traditional scalar quantization scheme; compared with the vector quantization algorithm, the method can greatly improve the quantization effect while only slightly reducing the operation rate, and has good performance.

[0166] In summary, compared with the existing scalar quantization and vector quantization technology, the method effectively reduces the number of bits of the signal fed back by the user end to the base station end in the downlink channel estimation process, and can obtain more accurate estimated channel state information matrix, and improves the data transmission rate of the system.

[0167] It should be noted that although the above describes the steps in a specific order, it does not mean that the steps must be performed in the above specific order, in fact, some of the steps can be executed concurrently, or even in a changed order, as long as the required function can be achieved.

[0168] The present application can be a system, a method, and / or a computer program product. The computer program product can include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present application.

[0169] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or punched tape, a magneto-optical storage device, or any suitable combination of the foregoing. A non-transitory, computer-readable storage medium does not include a signal.

[0170] Having described various embodiments of the application, it is to be understood that the above description is meant not to limit and not to encompass all of the possible embodiments covered by the claims. Many modifications and variations of this application can be apparent to those of ordinary skill in the art without departing from the spirit and scope of the described embodiments. It is intended that the scope of the application should only be limited by the appended claims.

Claims

1. A channel estimation method, applied at a base station, for estimating a downlink channel state, characterized in that: Methods include: S1. Obtain a vector quantization codebook preset for the MIMO system, which includes multiple code vectors for representing different pilot received signal vectors and an index value of each code vector; S2. Sending pilot signals to multiple user terminals, and receiving signals fed back by the multiple user terminals to the base station based on their pilot received signals, where the fed-back signals are index values ​​of code vectors most similar to the quantized vectors of their pilot received signals, as searched from the vector quantization codebook by the user terminals; S3. Obtain an estimated quantization vector of the pilot reception signal based on the vector quantization codebook and the signal fed back by each user terminal; S4. Perform channel estimation based on the estimated quantization vector using a preset estimation method to obtain an estimated downlink channel state information matrix corresponding to each user terminal, wherein the preset estimation method includes: With the optimization goal of minimizing the gap between the true downlink channel state information matrix and the estimated downlink channel state information matrix, an optimal post-processing matrix optimization function based on the minimum mean square error criterion is constructed; Obtaining an optimal post-processing matrix based on the optimization function and the estimated quantized vector of the pilot received signal; Obtaining an estimated downlink channel state information matrix according to the optimal post-processing matrix and the estimated quantized vector of the pilot reception signal; The optimal post-processing matrix optimization function is as follows: ， in, represents minimizing the gap between the true downlink channel state information matrix and the estimated downlink channel state information matrix, represents the post-processing matrix, represents the expected calculation symbol, represents the two-norm of the vector, , Indicates vectorizing matrix operations. represents the real downlink channel state information matrix, , represents the estimated downlink channel state information matrix, Indicates the pilot reception signal Known estimated value.

2. The method according to claim 1, characterized in that In the S1, the vector quantization codebook is generated by: Constructing a vector source training set for an indoor communication scenario, including obtaining multiple locations of the indoor communication scenario, calculating pilot received signal vectors of users at the multiple locations, and obtaining the vector source training set; According to the vector source training set and the preset number of quantization bits, a vector quantization algorithm is used to generate a vector quantization codebook, which includes multiple code vectors and an index value corresponding to each code vector. The preset number of quantization bits is the number of bits of the index value and is less than the number of bits of the code vector.

3. The method according to claim 2, characterized in that The method of constructing a vector source training set for indoor communication scenarios includes: Using the random walk model, multiple locations of users randomly walking in the predetermined room are generated; According to the transmitted pilot signal and the channel model, the pilot received signal vectors of multiple users wandering to various positions are calculated to obtain a vector source training set.

4. The method according to claim 2, characterized in that The vector quantization algorithm includes multiple rounds of iterations according to a preset number of quantization bits, and each round of iteration includes: Obtaining a code vector, wherein the first round uses a code vector calculated based on each sample and the number of samples in the vector source training set, and each subsequent round uses the code vector obtained in the previous round; Obtain a disturbance coefficient, and perform splitting calculation on the obtained code vector according to the disturbance coefficient to obtain multiple code vectors after splitting; Each code vector in the multiple code vectors after splitting is updated once or multiple times according to preset rules to obtain the code vector of the current round.

5. The method according to claim 4, characterized in that The preset rules include: Obtain multiple code vectors for the current time, use the multiple code vectors after splitting for the first time, and use the multiple code vectors after the previous update each time after the first time; According to the multiple code vectors of the current time, the vector source training set is divided into multiple subsets based on the nearest neighbor rule, and each code vector of the current time corresponds to a subset; The total distortion is calculated based on the multiple divided subsets, and whether the total distortion has converged is determined. If not, the multiple code vectors of the current round are updated to obtain multiple updated code vectors, including averaging the sum of all samples in the subset corresponding to each code vector to obtain the updated code vector. Otherwise, the above updating process is stopped to obtain the code vector of the current round.

6. The method according to claim 1, characterized in that In S2, the feedback signal is obtained by the user end in the following processing manner: Performing vector quantization on the pilot reception signal to obtain a quantized vector of the pilot reception signal; A code vector that is most similar to the quantized vector of the pilot received signal is searched from the vector quantization codebook, and an index value of the most similar code vector is used as a feedback signal.

7. The method according to claim 6, characterized in that The method of performing vector quantization on the pilot received signal is as follows: ， in, represents the quantized vector of its pilot received signal, Indicates vectorizing matrix operations. represents the pilot reception signal, Indicates sending a pilot signal. represents the tensor product, express The identity matrix, Indicates the number of receiving antennas of all user terminals, represents the real downlink channel state information matrix, Represents the noise matrix when receiving the pilot signal.

8. A communication system based on the method according to any one of claims 1 to 7, characterized in that: It includes multiple user terminals and base station terminals, among which, The base station is configured as follows: Obtaining a vector quantization codebook preset for the MIMO system, the codebook including a plurality of code vectors for representing different pilot received signal vectors and an index value of each code vector; Sending pilot signals to multiple user terminals and receiving signals fed back by the multiple user terminals based on their respective pilot received signals; Based on the signal fed back by each user terminal and the vector quantization codebook, an estimated quantization vector of the pilot reception signal is obtained; Performing channel estimation based on the estimated quantization vector using a preset estimation method to obtain an estimated downlink channel state information matrix corresponding to each user terminal; and Optimizing the communication parameter configuration of the base station according to the estimated downlink channel state information matrix corresponding to each user terminal, and sending data signals to each user terminal based on the optimized communication parameter configuration; Each client is configured as: Obtaining a vector quantization codebook preset for the MIMO system, searching the vector quantization codebook for an index value of a code vector that is most similar to the quantization vector of the pilot received signal based on the pilot received signal, and feeding back the index value of the most similar code vector as a signal to the base station; and Receives the data signal sent by the base station and obtains data based on its own location and the received data signal.

Images