CN114157331A - Large-scale MIMO channel state information feedback method based on pseudo-complex value input - Google Patents
Large-scale MIMO channel state information feedback method based on pseudo-complex value input Download PDFInfo
- Publication number
- CN114157331A CN114157331A CN202111563073.8A CN202111563073A CN114157331A CN 114157331 A CN114157331 A CN 114157331A CN 202111563073 A CN202111563073 A CN 202111563073A CN 114157331 A CN114157331 A CN 114157331A
- Authority
- CN
- China
- Prior art keywords
- layer
- convolution
- encoder
- decoder
- state information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 23
- 239000011159 matrix material Substances 0.000 claims abstract description 65
- 238000012549 training Methods 0.000 claims abstract description 14
- 230000006835 compression Effects 0.000 claims abstract description 12
- 238000007906 compression Methods 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 25
- 230000004913 activation Effects 0.000 claims description 12
- 239000013598 vector Substances 0.000 claims description 12
- 238000010586 diagram Methods 0.000 claims description 9
- 210000002569 neuron Anatomy 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000000137 annealing Methods 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000007493 shaping process Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000003213 activating effect Effects 0.000 claims description 2
- 238000011084 recovery Methods 0.000 abstract description 4
- 238000004891 communication Methods 0.000 description 4
- 238000004088 simulation Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000002679 ablation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/0413—MIMO systems
- H04B7/0417—Feedback systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/0413—MIMO systems
- H04B7/0456—Selection of precoding matrices or codebooks, e.g. using matrices antenna weighting
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/06—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station
- H04B7/0613—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission
- H04B7/0615—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission of weighted versions of same signal
- H04B7/0619—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission of weighted versions of same signal using feedback from receiving side
- H04B7/0658—Feedback reduction
- H04B7/0663—Feedback reduction using vector or matrix manipulations
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Abstract
The invention discloses a large-scale MIMO channel state information feedback method based on pseudo-complex value input, which comprises the following operation steps: (1) obtaining the channel state information of a space-frequency domain at a mobile user terminal and preprocessing the channel state information to obtain a channel matrix in an angular delay domain; (2) building a feedback model CLPNet comprising an encoder and a decoder; the encoder is deployed at a user side, and the decoder is deployed at a base station; the user side inputs the channel matrix into the encoder and feeds back the channel matrix to the base station; the base station obtains a reconstructed channel matrix according to the input to the decoder; (3) training a feedback model CLPNet to minimize a channel matrix and obtain model parameters; (4) the method and the device can rapidly and accurately improve the recovery precision and reduce the feedback overhead under the condition of low compression ratio.
Description
Technical Field
The invention relates to the technical field of communication, in particular to a large-scale MIMO channel state information feedback method.
Background
A large-scale Multiple Input Multiple Output (MIMO) technology is widely recognized as one of core technologies of next-generation communication systems. After a large number of antennas are configured at the base station end, the base station can greatly improve the channel capacity by using diversity reception and multiplexing technology. The precondition for obtaining the potential gain of massive MIMO is that the base station must receive accurate downlink channel state information. In Frequency Division Duplex (FDD) mode, where most cellular systems operate, there is no reciprocity between channels. Therefore, the device side (UE) must accurately feed back the downlink Channel State Information (CSI) to the Base Station (BS). However, when channel estimation is performed by using the pilot training downlink, the overhead increases exponentially with the increase of the number of antennas, which is a main obstacle for practical deployment of FDD massive MIMO systems. Therefore, CSI compression is required before feedback to reduce overhead.
In the current research on channel state information feedback of the massive MIMO system, the main idea of most algorithms is to reduce feedback overhead by using the space-time correlation of the massive MIMO system. Specifically, the channel matrix may be converted into a vector that may be sparsely represented using designed orthogonal bases. Then, random compression is carried out on the sparse vector at the user end to obtain a low-dimensional measurement value, the low-dimensional measurement value is transmitted to the base station end through a feedback link, and the base station reconstructs the original sparse vector by means of a compressive sensing theory. However, in practical situations, it is difficult to achieve an absolute sparseness effect by using the orthogonal basis. Therefore, the premise of compressed sensing is that it is assumed to be sparse and not much practical. And the traditional compressed sensing theory has the fatal defects of slow iteration, low efficiency and insufficient utilization of a channel structure.
In response to these problems, a massive MIMO channel state information feedback method based on deep learning is proposed as patent application No. 201810090648.0. The method treats a pure digital composition channel matrix as an image, but does not consider the physical characteristic that a signal is a complex number, most processing methods are to separate the complex number image into real and imaginary parts, and the phase characteristic of the complex number cannot be expressed, so that the original physical information is lost. Therefore, in most cases, the recovery accuracy is insufficient, and in particular, the recovery accuracy is excessively poor in the case of a low compression ratio. In the massive MIMO time-varying channel state information compression feedback and reconstruction method disclosed in patent No. 201810833351.9, although the method is deployed at the base station, the complexity is too high, and the performance is not improved too much.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a large-scale MIMO channel state information feedback method based on pseudo-complex value input, which can quickly and accurately improve the recovery precision and reduce the feedback overhead under the condition of low compression ratio.
The purpose of the invention is realized as follows: a large-scale MIMO channel state information feedback method based on pseudo-complex value input comprises the following operation steps:
(1) obtaining the channel state information of the space-frequency domain at the mobile user terminal and preprocessing the channel state information to obtain a channel matrix H in an angular delay domaina;
(2) Building a feedback model CLPNet comprising an encoder and a decoder; the encoder is deployed at a user side, and the decoder is deployed at a base station; user side will channel matrix HaInputting the code word c of the lower dimension into an encoder and feeding back the code word c to the base station; the code word c obtained by the base station is input to a decoder to obtain a reconstructed channel matrix
(3) Training the feedback model CLPNet to make the channel matrixAnd channel matrix HaMinimizing and obtaining model parameters;
(4) carrying out zero filling and two-dimensional inverse transformation operation on the matrix output by the feedback model CLPNet to recover to obtain the original matrixChannel matrix of starting space frequency domainAnd finishing feedback.
As a further limitation of the present invention, the step (2) comprises the steps of:
(21) channel matrix H of angle delay domainaThe encoder is used as the input of the encoder, the output is a compressed code word c, and the encoder sequentially comprises a first convolution layer, a space attention block, a second convolution layer and a full connection layer;
(22) feeding the code word c back to the base station end, decoding by a designed decoder, wherein the decoder takes the code word c as input and outputs the code word c and a channel matrix H through zero filling operation of a convolution kernelaChannel matrix of the same dimensionThe decoder includes a fully-connected layer, two convolutional layers and an MLP block.
As a further limitation of the present invention, the encoder inputs two layers of real matrixes of 32 × 32 size, the two layers are split of real part and imaginary part of complex matrix H, the first layer of the encoder is a convolutional layer with 16 channel output, the convolutional kernel size is 1 × 1, and the function is to combine two layers of matrices separated by virtual and real into one feature diagram, thus completing real and virtual fusion and retaining certain physical information; the second layer of the encoder is a spatial attention module, and the operation inside the module is to generate two 2D feature maps by adopting average pooling and maximum pooling operations on a channel C of an input F; then splicing the two feature maps to generate a compressed spatial characteristic descriptor, performing convolution operation on the compressed spatial characteristic descriptor and a standard layer, wherein the size of a convolution kernel is 3, the output is 1 channel, generating a 2D spatial attention mask, activating the mask by using sigmoid, and finally performing convolution operation on the 2D spatial attention mask and the original feature map FiMultiplying to obtain F with spatial attentiono:Fo=Fi(σ(fc(Favg;Fmax) ); the output of the space attention block is connected with a convolution kernel of 3 multiplied by 3, so that the channel is restored into two channels; reducing the two matrixes into a one-dimensional matrix with the sizeThe characteristic diagram is 2048 multiplied by 1, is input into a full connection layer containing M neurons, and a linear activation function is adopted to output a vector with the size of M multiplied by 1, namely a code word c which is transmitted to a base station by a user after being compressed; the compression ratio CR can be changed by changing the dimension M of the output layer, where CR is M/N, and N is the number of neurons in the input layer.
As a further limitation of the present invention, the first layer of the decoder is a fully-connected layer whose output contains 2048 neurons, takes the received codeword c as input, takes a matrix of size mx 1, and uses a linear activation function to output a vector of 2048 × 1; recombining the vectors into two layers of matrixes with the size of 32 multiplied by 32 through shaping reshape operation, inputting the matrixes into a convolution layer of a second layer, wherein the convolution belongs to head convolution operation, the convolution kernel size is 5 multiplied by 5, and the output size is 32 channels; the third layer is a CLPblock unit which is divided into a main line and a sub-line, the main line is composed of three convolution layers with different convolution kernel sizes in parallel, the convolution kernel sizes are distributed to 3 × 3, 5 × 5 and 7 × 7, a batch normalization and linear correction unit LaekyRelu with leakage is used in the convolution layers, and 96 channels are reduced to 32 channels through 1 × 1 convolution kernels after splicing; finally, according to the residual error idea, a sub-line is led out from the last convolution layer and is added with the output of the 1 multiplied by 1 convolution layer, and a LeakyRelu function is adopted for activation; the second convolutional layer has MLPblock as input and two 32 × 32 matrix layers as output, and is activated by HardSigmoid function, and the output of the activation function is used as the final reconstructedReal and imaginary parts of (c).
As a further limitation of the present invention, the model parameters in step (3) include weight, offset of the fully-connected layer, and convolution kernel, offset of the convolutional layer.
As a further limitation of the present invention, in the step (3), Adam optimization mode and end-to-end learning mode are adopted, the learning rate adopts dynamic learning rate, namely cosine annealing algorithm, the encoder and the decoder are jointly trained to minimize the cost function, and the cost function of the whole CLPNet framework is designed as the channel matrix output by the decoderAnd the true channel matrix HaThe cost function is described as follows:
wherein C is the number of all samples in the training set, | · | | purple2Is the Euclidean norm;fe,fd,θeand thetadRespectively, an encoder, a decoder, encoder parameters and decoder parameters.
Compared with the prior art, the invention has the beneficial effects that:
compared with CsiNet which takes a channel matrix as an image to carry out virtual and real separate processing, CLPNet not only reserves the phase characteristic, but also distributes different weights to different clusters by utilizing the characteristic of the space attention mechanism, and gives more attention to distinguishable paths with specific delay and arrival angles; the decoder end designs a parallel large convolution channel and a plurality of convolution kernels, so that the received image has more characteristics and different receptive fields; under the same compression ratio. CLPNet can obtain smaller reconstruction error, thereby improving the quality of the reconstructed matrix.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flow chart of the present invention.
FIG. 2 is a diagram of the overall architecture of the CLPNet according to the present invention.
FIG. 3 is a diagram of a spatial attention unit according to the present invention.
Fig. 4 is a diagram of a CLPblock unit of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Examples
Fig. 1 shows a massive MIMO channel state information feedback method based on pseudo-complex value input, which includes the following steps:
(1) obtaining the channel state information of the space-frequency domain at the mobile user terminal and preprocessing the channel state information to obtain a channel matrix H in an angular delay domaina(ii) a In this embodiment, in a large-scale MIMO communication system, FDD is used, and N is provided at the base stationtThe system uses the carrier modulation mode of orthogonal frequency division multiplexing, the number of sub-carriers is as follows150000 channel matrix samples of the air frequency domain are produced by a COST2100 channel model through Matlab simulation software in an indoor environment of 5.3GHz, and the samples are sequentially divided into a training set (100000), a verification set (30000) and a test set (20000) according to a certain proportion; the produced matrix is directly used as the input of the model, and the user terminal estimates the downlink channel state information of the large-scale MIMO system according to the training sequence or the pilot frequency sequence sent by the base station to obtain the space-frequency domain channel matrixThe matrix size isSpace-frequency domain channel matrixTwo-dimensional discrete Fourier transform is carried out to obtain a channel sparse matrix H of an angular delay domain, the transformation is completed through two DFT matrixes, and the expression of a coefficient matrix isWhereinAndsince the multipath delay is finite, the channel matrix can be advanced in the angular delay domainRow by row, resulting in a matrix HaThus, too much information loss is not caused; in this example, 32 rows are taken, i.e. Na32, sub-carrierTaking 1024; the matrix H thus finally corrected is based on the number of transmit antennas and the number of rows retainedaThe size is 32 × 32;
(2) building a feedback model CLPNet comprising an encoder and a decoder as shown in FIG. 2;
the encoder is deployed at a user end and comprises two convolutional layers, a space attention block and a full connection layer, the input of the encoder is two real matrixes with the size of 32 multiplied by 32, the two real matrixes and the imaginary matrix are split, the first layer is a convolutional layer with the output of 16 channels, the size of a convolutional kernel is 1 multiplied by 1, the function of the encoder is to combine two matrixes with virtual and real parts separated into a characteristic diagram, so that real and virtual fusion is completed, and certain physical information is reserved; the second level is a spatial attention module as shown in FIG. 3, the operation inside the module is to employ average pooling on the channel C of the input FAnd max pooling operation, generating two 2D feature maps, havingThe two feature maps are then stitched to generate a compressed spatial property descriptorThen the data is convolved with a standard layer (the convolution kernel size is 3), the output is 1 channel, and a 2D space attention mask is generatedActivating a mask by using sigmoid, and finally matching the mask with an original feature map FiMultiplying to obtain F with spatial attentiono:Fo=Fi(σ(fc(Favg;Fmax) ); the output of the space attention block is connected with a convolution kernel of 3 multiplied by 3, so that the channel is restored into two channels; reducing the two matrixes into one-dimensional matrixes, inputting characteristic diagrams with the matrix size of 2048 multiplied by 1 into a full-connection layer containing M neurons, and outputting vectors with the size of M multiplied by 1 by adopting a linear activation function, namely the code words c which are compressed by a base station and are transmitted by a user; the compression ratio CR can be changed by changing the dimension M of the output layer, wherein CR is M/N, and N is the number of neurons of the input layer;
the decoder is deployed at the base station end; the first layer outputs a fully-connected layer containing 2048 neurons, takes an accepted code word c as input, adopts a matrix with the size of M multiplied by 1, adopts a linear activation function, and outputs a vector of 2048 multiplied by 1; through shaping reshape operation, the vectors are recombined into two layers of matrixes with the size of 32 multiplied by 32 and input into a convolution layer of a second layer, the convolution belongs to head convolution operation, the performance is highest when the convolution kernel size is 5 multiplied by 5 through ablation verification, the output size is 32 channels, and more convolution channels can obtain more characteristic maps; the third layer is a CLPblock unit as shown in FIG. 4, which is divided into a main line and a sub-line, wherein the main line is composed of three convolution layers with different convolution kernel sizes in parallel, the convolution kernel sizes are distributed to 3 × 3, 5 × 5 and 7 × 7, and the convolution layers are divided into three sub-linesBy decomposing the large convolution kernel into multiple 3 x 3 forms, 5 x 5 can be replaced by two 3 x 3 convolution kernels, and 7 x 7 can be replaced by three 3 x 3 convolution kernels, which effectively reduces the parameters; batch normalization and a linear correction unit LaekyRelu with leakage are used in the convolutional layer, and 96 channels are reduced to 32 channels through a 1 × 1 convolution kernel after splicing; finally, according to the residual error idea, a sub-line is led out from the last convolution layer and is added with the output of the 1 multiplied by 1 convolution layer, and a LeakyRelu function is adopted for activation; the second convolutional layer has MLPblock as input and two 32 × 32 matrix layers as output, and is activated by HardSigmoid function, and the output of the activation function is used as the final reconstructedReal and imaginary parts of (c).
(3) Training the feedback model CLPNet to make the channel matrixAnd channel matrix HaMinimizing and obtaining model parameters; designing cost function of whole CLPNet structure as channel matrix of decoder outputAnd the true channel matrix HaMean square error of, i.e. cost functionWherein C is the number of all samples in the training set, | · | | purple2Is the Euclidean norm;wherein f ise,fd,θeAnd thetadRespectively representing an encoder, a decoder, encoder parameters and decoder parameters;
training is carried out in an end-to-end learning mode according to a channel matrix generated by COST2100, an optimizer in the training process selects Adam for optimization, 200 samples are obtained in each batch, the learning rate adopts a cosine annealing algorithm, the learning rate is from the initial 0.002 to the final 0.00005, and the whole training set is traversed for 1000 times in the mode; the model with the best selectivity of the available verification set is traversed every 10 times in the training process, and the test set is used for testing
(4) The trained CLPNet model is finally deployed at a user side and a base station for channel state information feedback of large-scale MIMO (multiple input multiple output), and the channel state information feedback is realized by reconstructing a matrixZero filling operation is carried out, two-dimensional discrete Fourier transform is carried out, and the channel matrix of the original space-frequency domain is obtained
The technical effects of the present invention are described below by experiments:
the verification example is a large-scale MIMO channel state information feedback method based on pseudo-complex value input, and a coder and a decoder framework are designed through data drive; compared with the simulation of the prior art, the invention adopts a sample generated by a COST2100MIMO channel in a 5.3GHz indoor scene, the transmitting antenna of a base station end is 32, a user end has a single antenna, the system adopts a carrier modulation mode of orthogonal frequency division multiplexing, the number of subcarriers is 1024, and the parameter setting is the default parameter selection of the invention.
Simulation content: under the condition of different compression ratios, the CLPNet and the CsiNet and CsiNet-LSTM in the prior art are applied to the reconstructed channel matrixAnd the original channel matrix HaCarrying out simulation comparison on the direct normalized mean square error; as shown in table one, the formula for normalizing the mean square error is described as follows:the unit is dB.
Watch 1
As can be seen from the table I, the error performance of the CLPNet under a plurality of compression ratios in the three simulation methods is greatly improved. The feasibility of the attention mechanism in the communication direction is also proved due to the fact that the weight distribution of the attention mechanism on a resolvable path is larger at the encoder end, and more characteristic graphs are generated by a large convolution channel at the decoder end. Therefore, CLPNet has potential application value in the operation of FDD massive MIMO channel state information feedback
The above description of the embodiments is only intended to facilitate the understanding of the method of the invention and its core idea. It should be noted that, for those skilled in the art, it is possible to make various improvements and modifications to the present invention without departing from the principle of the present invention, and those improvements and modifications also fall within the scope of the claims of the present invention.
Claims (6)
1. A large-scale MIMO channel state information feedback method based on pseudo-complex value input is characterized by comprising the following operation steps:
(1) obtaining the channel state information of the space-frequency domain at the mobile user terminal and preprocessing the channel state information to obtain a channel matrix H in an angular delay domaina;
(2) Building a feedback model CLPNet comprising an encoder and a decoder; the encoder is deployed at a user side, and the decoder is deployed at a base station; user side will channel matrix HaInputting the code word c of the lower dimension into an encoder and feeding back the code word c to the base station; the code word c obtained by the base station is input to a decoder to obtain a reconstructed channel matrix
(3) Training the feedback model CLPNet to make the channel matrixAnd channel matrix HaMinimizing and obtaining model parameters;
2. The feedback method of massive MIMO channel state information based on pseudo-complex value input according to claim 1, wherein the step (2) comprises the steps of:
(21) channel matrix H of angle delay domainaThe encoder is used as the input of the encoder, the output is a compressed code word c, and the encoder sequentially comprises a first convolution layer, a space attention block, a second convolution layer and a full connection layer;
(22) feeding the code word c back to the base station end, decoding by a designed decoder, wherein the decoder takes the code word c as input and outputs the code word c and a channel matrix H through zero filling operation of a convolution kernelaChannel matrix of the same dimensionThe decoder includes a fully-connected layer, two convolutional layers and an MLP block.
3. The feedback method of the large-scale MIMO channel state information based on pseudo-complex value input according to claim 2, wherein the encoder input is two layers of real matrices of 32 × 32 size, the two layers are the real and imaginary parts split of the complex matrix H, the first layer of the encoder is a convolutional layer of which the output is 16 channels, the convolutional kernel size is 1 × 1, and the function is to combine the two layers of matrices separated by virtual and real into one characteristic diagram, thus completing the real and virtual fusion and retaining certain physical information; the second layer of the encoder is a spatial attention module, and the operation inside the module is to generate two 2D feature maps by adopting average pooling and maximum pooling operations on a channel C of an input F; then splicing the two feature maps to generate a compressed spatial characteristic descriptor, performing convolution operation on the compressed spatial characteristic descriptor and a standard layer, wherein the size of a convolution kernel is 3, the output is 1 channel, generating a 2D spatial attention mask, activating the mask by using sigmoid, and finally performing convolution operation on the 2D spatial attention mask and the original feature map FiMultiplying to obtain F with spatial attentiono:Fo=Fi(σ(fc(Favg;Fmax) ); the output of the space attention block is connected with a convolution kernel of 3 multiplied by 3, so that the channel is restored into two channels; reducing the two matrixes into one-dimensional matrixes, inputting characteristic diagrams with the matrix size of 2048 multiplied by 1 into a full-connection layer containing M neurons, and outputting vectors with the size of M multiplied by 1 by adopting a linear activation function, namely the code words c which are compressed by a base station and are transmitted by a user; the compression ratio CR can be changed by changing the dimension M of the output layer, where CR is M/N, and N is the number of neurons in the input layer.
4. The feedback method of channel state information of massive MIMO based on pseudo-complex value input as claimed in claim 3 wherein, the first layer of the decoder is a fully connected layer outputting 2048 neurons, taking the received codeword c as input, a matrix of size M x 1, using linear activation function, outputting a vector of 2048 x 1; recombining the vectors into two layers of matrixes with the size of 32 multiplied by 32 through shaping reshape operation, inputting the matrixes into a convolution layer of a second layer, wherein the convolution belongs to head convolution operation, the convolution kernel size is 5 multiplied by 5, and the output size is 32 channels; the third layer is a CLPblock unit which is divided into a main line and a sub-line, the main line is composed of three convolution layers with different convolution kernel sizes in parallel, the convolution kernel sizes are distributed to 3 × 3, 5 × 5 and 7 × 7, a batch normalization and linear correction unit LaekyRelu with leakage is used in the convolution layers, and 96 channels are reduced to 32 channels through 1 × 1 convolution kernels after splicing; finally, according to the residual error idea, a sub-line is led out from the last convolution layer and is added with the output of the 1 multiplied by 1 convolution layer, and a LeakyRelu function is adopted for activation; the second convolutional layer has MLPblock as input and two 32 × 32 matrix layers as output, and is activated by HardSigmoid function, and the output of the activation function is used as the final reconstructedReal and imaginary parts of (c).
5. The feedback method of massive MIMO channel state information based on pseudo-complex value input of claim 4, wherein the model parameters in step (3) comprise weight, bias of full link layer and convolution kernel, bias of convolution layer.
6. The feedback method of the large-scale MIMO channel state information based on the pseudo-complex value input according to claim 5, wherein the step (3) adopts an Adam optimization mode and an end-to-end learning mode, the learning rate adopts a dynamic learning rate (cosine annealing algorithm), an encoder and a decoder are jointly trained to minimize a cost function, and the cost function of the whole CLPNet framework is designed to be a channel matrix output by the decoderAnd the true channel matrix HaThe cost function is described as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111563073.8A CN114157331B (en) | 2021-12-20 | 2021-12-20 | Large-scale MIMO channel state information feedback method based on pseudo complex value input |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111563073.8A CN114157331B (en) | 2021-12-20 | 2021-12-20 | Large-scale MIMO channel state information feedback method based on pseudo complex value input |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114157331A true CN114157331A (en) | 2022-03-08 |
CN114157331B CN114157331B (en) | 2023-11-24 |
Family
ID=80451667
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111563073.8A Active CN114157331B (en) | 2021-12-20 | 2021-12-20 | Large-scale MIMO channel state information feedback method based on pseudo complex value input |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114157331B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114978264A (en) * | 2022-06-29 | 2022-08-30 | 内蒙古大学 | Hybrid precoding method based on terahertz MIMO system |
CN116505989A (en) * | 2023-05-06 | 2023-07-28 | 四川农业大学 | CSI feedback method and system based on complex encoder hybrid neural network |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108390706A (en) * | 2018-01-30 | 2018-08-10 | 东南大学 | A kind of extensive mimo channel state information feedback method based on deep learning |
CN109672464A (en) * | 2018-12-13 | 2019-04-23 | 西安电子科技大学 | Extensive mimo channel state information feedback method based on FCFNN |
CN110311718A (en) * | 2019-07-05 | 2019-10-08 | 东南大学 | Quantization and inverse quantization method in a kind of extensive mimo channel status information feedback |
CN110350958A (en) * | 2019-06-13 | 2019-10-18 | 东南大学 | A kind of more multiplying power compressed feedback methods of CSI of extensive MIMO neural network based |
US20200220593A1 (en) * | 2019-01-04 | 2020-07-09 | Industrial Technology Research Institute | Communication system and codec method based on deep learning and known channel state information |
CN111555781A (en) * | 2020-04-27 | 2020-08-18 | 天津大学 | Large-scale MIMO channel state information compression and reconstruction method based on deep learning attention mechanism |
-
2021
- 2021-12-20 CN CN202111563073.8A patent/CN114157331B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108390706A (en) * | 2018-01-30 | 2018-08-10 | 东南大学 | A kind of extensive mimo channel state information feedback method based on deep learning |
CN109672464A (en) * | 2018-12-13 | 2019-04-23 | 西安电子科技大学 | Extensive mimo channel state information feedback method based on FCFNN |
US20200220593A1 (en) * | 2019-01-04 | 2020-07-09 | Industrial Technology Research Institute | Communication system and codec method based on deep learning and known channel state information |
CN110350958A (en) * | 2019-06-13 | 2019-10-18 | 东南大学 | A kind of more multiplying power compressed feedback methods of CSI of extensive MIMO neural network based |
CN110311718A (en) * | 2019-07-05 | 2019-10-08 | 东南大学 | Quantization and inverse quantization method in a kind of extensive mimo channel status information feedback |
CN111555781A (en) * | 2020-04-27 | 2020-08-18 | 天津大学 | Large-scale MIMO channel state information compression and reconstruction method based on deep learning attention mechanism |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114978264A (en) * | 2022-06-29 | 2022-08-30 | 内蒙古大学 | Hybrid precoding method based on terahertz MIMO system |
CN114978264B (en) * | 2022-06-29 | 2023-07-25 | 内蒙古大学 | Mixed precoding method based on terahertz MIMO system |
CN116505989A (en) * | 2023-05-06 | 2023-07-28 | 四川农业大学 | CSI feedback method and system based on complex encoder hybrid neural network |
CN116505989B (en) * | 2023-05-06 | 2024-02-06 | 四川农业大学 | CSI feedback method and system based on complex encoder hybrid neural network |
Also Published As
Publication number | Publication date |
---|---|
CN114157331B (en) | 2023-11-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112737985B (en) | Large-scale MIMO channel joint estimation and feedback method based on deep learning | |
CN109672464B (en) | FCFNN-based large-scale MIMO channel state information feedback method | |
CN108390706B (en) | Large-scale MIMO channel state information feedback method based on deep learning | |
CN111698182B (en) | Time-frequency blocking sparse channel estimation method based on compressed sensing | |
Cao et al. | Lightweight convolutional neural networks for CSI feedback in massive MIMO | |
CN109257309B (en) | High-performance large-scale MIMO downlink transmission channel estimation method | |
CN111181671B (en) | Deep learning-based downlink channel rapid reconstruction method | |
TWI294723B (en) | Compact feedback for closed loop mimo systems | |
CN114157331B (en) | Large-scale MIMO channel state information feedback method based on pseudo complex value input | |
CN111555781B (en) | Large-scale MIMO channel state information compression and reconstruction method based on deep learning attention mechanism | |
CN113748614A (en) | Channel estimation model training method and device | |
CN109327918B (en) | Low-overhead FDD large-scale MIMO downlink channel reconstruction method | |
CN110289898A (en) | A kind of channel feedback method based on the perception of 1 bit compression in extensive mimo system | |
CN105978674A (en) | FDD large-scale MIMO channel estimation pilot frequency optimization method based on compressed sensing | |
Elbir et al. | Federated learning for physical layer design | |
Yin et al. | Deep CSI compression for massive MIMO: A self-information model-driven neural network | |
CN106788631B (en) | Large-scale MIMO reciprocity calibration method based on local calibration | |
Guo et al. | Deep learning for joint channel estimation and feedback in massive MIMO systems | |
CN115865145A (en) | Large-scale MIMO channel state information feedback method based on Transformer | |
CN115208442A (en) | Two-stage beam training method of ultra-large scale array communication system | |
CN116248156A (en) | Deep learning-based large-scale MIMO channel state information feedback and reconstruction method | |
CN106130938B (en) | Multi-user joint channel estimation method for TDD large-scale MIMO system | |
CN114567359A (en) | CSI feedback method based on multi-resolution fusion convolution feedback network in large-scale MIMO system | |
CN114884549A (en) | Large-scale MIMO channel state information feedback method based on deep learning | |
CN114884775A (en) | Deep learning-based large-scale MIMO system channel estimation method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |