WO2022256988A1 - 信息反馈方法、装置、用户设备、基站、系统模型及存储介质 - Google Patents

信息反馈方法、装置、用户设备、基站、系统模型及存储介质 Download PDF

Info

Publication number
WO2022256988A1
WO2022256988A1 PCT/CN2021/098711 CN2021098711W WO2022256988A1 WO 2022256988 A1 WO2022256988 A1 WO 2022256988A1 CN 2021098711 W CN2021098711 W CN 2021098711W WO 2022256988 A1 WO2022256988 A1 WO 2022256988A1
Authority
WO
WIPO (PCT)
Prior art keywords
matrix
information
csi
information matrix
image
Prior art date
Application number
PCT/CN2021/098711
Other languages
English (en)
French (fr)
Inventor
陈栋
池连刚
Original Assignee
北京小米移动软件有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 北京小米移动软件有限公司 filed Critical 北京小米移动软件有限公司
Priority to PCT/CN2021/098711 priority Critical patent/WO2022256988A1/zh
Priority to EP21944499.9A priority patent/EP4354755A1/en
Priority to CN202180001750.9A priority patent/CN115917982A/zh
Publication of WO2022256988A1 publication Critical patent/WO2022256988A1/zh

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B7/00Radio transmission systems, i.e. using radiation field
    • H04B7/02Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
    • H04B7/04Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
    • H04B7/0413MIMO systems
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B7/00Radio transmission systems, i.e. using radiation field
    • H04B7/02Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
    • H04B7/04Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
    • H04B7/06Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station

Definitions

  • the present disclosure relates to the technical field of communications, and in particular to an information feedback method, device, user equipment, base station, system model, and storage medium.
  • m-MIMO massive Multiple-Input Multiple-Output, large-scale multiple-input multiple-output
  • CSI Channel State Information
  • the method for the UE to feed back the CSI information matrix mainly includes:
  • Method 1 Use the AoD (Angle-of-departure, offset angle) adaptive subspace codebook to quantize the CSI information matrix into a certain number of bits, and feed it back to the base station, so that the base station can reconstruct the CSI based on the certain number of bits information matrix.
  • AoD Angle-of-departure, offset angle
  • Method 2 Transform the CSI information matrix into a sparse matrix in the channel space, and randomly compress and sample the sparse matrix to obtain low-dimensional measurement values, and feed them back to the base station, so that the base station reconstructs the CSI information matrix based on the low-dimensional measurement values.
  • Method 3 The neural network model based on DL (Deep Learning, deep learning) feeds back the CSI information matrix to the base station.
  • the first method when the number of bits to be converted by the CSI information matrix is large, the overhead is still relatively large.
  • the low-latitude measurement values are randomly sampled, and the structural characteristics of the CSI information matrix are not considered, so that the obtained low-latitude measurement values cannot accurately reflect the original CSI information matrix, so that the CSI information matrix reconstructed by the base station There is a large difference from the original CSI information matrix sent by the UE, resulting in low reconstruction accuracy of the base station.
  • the third method the complexity of training the neural network model is high, and the convergence is slow. At the same time, the accuracy of base station reconstruction is also low.
  • the information feedback method, device, user equipment, base station, system model and storage medium proposed in the present disclosure are used to solve the technical problems of low base station reconstruction accuracy and high cost in the existing information feedback method.
  • the information feedback method proposed in the embodiment of the first aspect of the present disclosure is applied to the UE, including:
  • the information feedback method proposed in the embodiment of the second aspect of the present disclosure is applied to a base station, including:
  • the first obtaining module is used to obtain the channel state information CSI information matrix
  • a screening module configured to screen elements in the CSI information matrix based on the amount of self-information of the CSI information matrix to obtain a sparse CSI information matrix
  • a feature encoder configured to determine feedback information based on elements in the sparse CSI information matrix, and send the feedback information to the base station.
  • a second acquiring module configured to acquire feedback information sent by the UE, and determine a preliminary CSI information matrix based on the feedback information
  • a reconstruction module configured to reconstruct a CSI information matrix based on the prepared CSI information matrix.
  • a user equipment provided in an embodiment of the fifth aspect of the present disclosure includes: a transceiver; a memory; and a processor, which are respectively connected to the transceiver and the memory, and configured to execute computer-executable instructions on the memory, The wireless signal transmission and reception of the transceiver is controlled, and the method proposed in the embodiment of the first aspect above can be implemented.
  • a base station which includes: a transceiver; a memory; and a processor connected to the transceiver and the memory respectively, configured to execute computer-executable instructions on the memory , controlling the wireless signal transmission and reception of the transceiver, and implementing the method proposed in the embodiment of the second aspect above.
  • An information feedback system model provided by an embodiment of the seventh aspect of the present disclosure, the system model at least includes the information feedback device described in the third aspect and the fourth aspect above.
  • the embodiment of the eighth aspect of the present disclosure proposes a training method for the information feedback system model described in the seventh aspect, including: performing the method described in the first aspect and/or the second aspect to train the The information feedback system model is trained.
  • the computer storage medium provided by the embodiment, wherein the computer storage medium stores computer-executable instructions; after the computer-executable instructions are executed by a processor, the method as described above can be implemented.
  • the UE will filter the elements in the CSI information matrix to obtain the sparse CSI information matrix. Specifically, when determining the sparse CSI information matrix, it will first The CSI image information matrix corresponding to the information matrix is divided into multiple sub-image information matrices, and then the distribution function corresponding to each sub-image information matrix is determined according to each sub-image information matrix and adjacent sub-image information matrices, and then determined based on the distribution function The self-information corresponding to each sub-image information matrix is obtained, and finally, the redundant information is screened based on the self-information to determine the sparse CSI information matrix. And, further, the UE determines feedback information according to the determined elements in the sparse CSI information matrix, and sends the feedback information to the base station, so that the base station reconstructs the CSI information matrix based on the feedback information.
  • the sparse CSI information matrix in the embodiments of the present disclosure is a matrix after redundant information is deleted, the convenience of subsequent operations on the sparse CSI information matrix can be ensured, and overhead can be reduced.
  • the feedback information determined by the UE includes the position information of the compressed elements in the sparse CSI information matrix, so that the subsequent base station can accurately reconstruct the
  • the CSI information matrix further ensures the reconstruction accuracy of the CSI information matrix.
  • FIG. 1 is a schematic flowchart of an information feedback method provided by an embodiment of the present disclosure
  • FIG. 2 is a schematic flow diagram of obtaining a sparse CSI information matrix provided by an embodiment of the present disclosure
  • FIG. 3 is a schematic flow diagram of determining a self-information matrix corresponding to a CSI image information matrix by using a second convolutional layer provided by an embodiment of the present disclosure
  • FIG. 4 is a schematic flowchart of determining feedback information provided by an embodiment of the present disclosure
  • FIG. 5 is a schematic flowchart of an information feedback method provided by an embodiment of the present disclosure.
  • FIG. 6 is a schematic structural diagram of an information feedback device provided by an embodiment of the present disclosure.
  • FIG. 7 is a schematic structural diagram of an information feedback device provided by an embodiment of the present disclosure.
  • FIG. 8 is a schematic structural diagram of an information feedback system model provided by an embodiment of the present disclosure.
  • Fig. 9 is a block diagram of a user equipment provided by an embodiment of the present disclosure.
  • Fig. 10 is a block diagram of a base station provided by an embodiment of the present disclosure.
  • first, second, third, etc. may use the terms first, second, third, etc. to describe various information, the information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of the embodiments of the present disclosure, first information may also be called second information, and similarly, second information may also be called first information.
  • first information may also be called second information
  • second information may also be called first information.
  • the words "if” and "if” as used herein may be interpreted as “at” or "when” or "in response to a determination.”
  • FIG. 1 is a schematic flowchart of an information feedback method provided by an embodiment of the present disclosure, which is applied to a UE. As shown in FIG. 1, the information feedback method may include the following steps:
  • Step 101 Obtain a CSI information matrix.
  • a UE may be a device that provides voice and/or data connectivity to a user.
  • UE can communicate with one or more core networks via RAN (Radio Access Network, wireless access network).
  • RAN Radio Access Network, wireless access network
  • UE can be an Internet of Things terminal, such as a sensor device, a mobile phone (or called a "cellular" phone) and a device with an Internet of Things
  • the computer of the terminal for example, may be a fixed, portable, pocket, hand-held, computer-built-in or vehicle-mounted device.
  • station Station, STA
  • subscriber unit subscriber unit
  • subscriber station subscriber station
  • mobile station mobile station
  • mobile station mobile
  • remote station remote station
  • access point remote terminal
  • user terminal or user agent.
  • the UE may also be a device of an unmanned aerial vehicle.
  • the UE may also be a vehicle-mounted device, for example, it may be a trip computer with a wireless communication function, or a wireless terminal connected externally to the trip computer.
  • the UE may also be a roadside device, for example, it may be a street lamp, a signal lamp, or other roadside devices with a wireless communication function.
  • a single antenna may be configured on a UE side, and multiple antennas may be configured on a base station side.
  • ULA Uniform Linear Array, Uniform Linear Array
  • the UE acquiring the CSI information matrix may include: the UE acquiring the CSI information matrix corresponding to each antenna channel of the base station.
  • the method for the UE to obtain the CSI information matrix may include: the UE determines the original CSI information matrix according to the pilot information sent by the base station, and then converts the obtained original CSI information matrix from the space frequency domain to the angular delay domain to obtain the CSI information matrix in the angular delay domain.
  • OFDM Orthogonal Frequency Division Multiplexing, Orthogonal Frequency Division Multiplexing
  • the size of the CSI information matrix H in the angular delay domain may be d 1 ⁇ N c ⁇ N t , where d 1 is the number of antenna channels, and N c indicates the antenna set by the base station number.
  • the CSI information matrix H in the angular delay domain may include a real part matrix H re and an imaginary part matrix H im .
  • the subsequent processing of the CSI information matrix H in the angular delay domain is specifically for the CSI
  • the real part matrix H re and the imaginary part matrix H im of the information matrix H are processed.
  • the above-mentioned scheme may also include:
  • Step 102 Filter the elements in the CSI information matrix based on the self-information amount of the CSI information matrix to obtain a sparse CSI information matrix.
  • step 101 may be implemented alone or together with step 102 . That is: any device can only execute step 101 to acquire the CSI information matrix. Any device can also perform steps 101 and 102 to obtain a CSI information matrix, and obtain a sparse CSI information matrix based on the CSI information matrix.
  • FIG. 2 is a schematic flow diagram of obtaining a sparse CSI information matrix provided by an embodiment of the present disclosure. As shown in FIG. 2, the method may include:
  • Step 201 mapping the CSI information matrix into a CSI image information matrix.
  • the CSI information matrix can be acquired in any appropriate way; that is, the CSI information matrix can be acquired in the same way as the aforementioned step 101, or can be acquired in any other feasible way .
  • mapping the CSI information matrix to a CSI image information matrix may include: converting the real part Hre of the CSI information matrix H into a CSI image information matrix corresponding to the real part Hre, and converting the CSI information matrix H The imaginary part Him of is transformed into the CSI image information matrix corresponding to the imaginary part Him.
  • Step 202 Use the first convolutional layer to perform feature processing on the CSI image information matrix to obtain n first feature maps, where n is an integer and n ⁇ 1.
  • the first convolutional layer may be a convolutional layer with gradient update.
  • the size of the convolution kernel of the first convolutional layer may be n ⁇ c ⁇ s ⁇ s, and the convolution step size may be L, where L ⁇ s, n, c, s, L are all positive integers.
  • the size of the convolution kernel of the first convolution layer may be 64 ⁇ 2 ⁇ 3 ⁇ 3, and the convolution step may be 1.
  • the n first feature maps may be n-dimensional feature maps corresponding to the CSI image information matrix.
  • n when n is equal to 64, it means: use the first convolutional layer to perform feature processing on the CSI image information matrix to obtain the 64-dimensional first feature corresponding to the CSI image information matrix picture.
  • Step 203 using the second convolutional layer to determine a self-information matrix corresponding to the CSI image information matrix, and the size of the self-information matrix is the same as that of the CSI information matrix.
  • FIG. 3 is a schematic flow diagram of using the second convolutional layer to determine the self-information matrix corresponding to the CSI image information matrix provided by an embodiment of the present disclosure. As shown in FIG. 3 , the method may include
  • Step 301 Add 0 to the periphery of the CSI image information matrix to expand the CSI image information matrix to obtain an extended CSI image information matrix, and divide the expanded CSI image information matrix into m sub-image matrices; wherein, m is an integer, m ⁇ 2, and m It is consistent with the number of elements included in the CSI image information matrix.
  • the CSI image information matrix can be obtained in any appropriate way; that is, the CSI image information matrix can be obtained in the same manner as the aforementioned step 201, or any other feasible method can be used to obtain the CSI information matrix.
  • the method for dividing the extended CSI image information matrix into m sub-image matrices may include: dividing the extended CSI image information matrix by using an x ⁇ s grid, where x, s Both are positive integers, and x and s can be the same or different.
  • the extended CSI image information matrix may be divided by a 7 ⁇ 7 grid.
  • Step 302 use the second convolutional layer to calculate the distribution function of each sub-image matrix based on each sub-image matrix and the adjacent sub-image matrices of each sub-image matrix, calculate the self-information amount of each sub-image matrix based on the distribution function, and based on The self-information of each sub-image matrix forms a self-information matrix.
  • the second convolutional layer may be a convolutional layer without gradient update.
  • the method of using the second convolutional layer to calculate the distribution function of each sub-image matrix based on each sub-image matrix and the adjacent sub-image matrices of each sub-image matrix may include the following steps:
  • Step a Select the ith sub-image matrix p i input to the second convolutional layer, where i is an integer and i ⁇ 1.
  • Step b Determine the adjacent sub-image matrix p i ′ of p i to obtain the adjacent sub-image set Z i corresponding to p i .
  • the adjacent sub-image set Z i may be determined based on the adjacent same distribution principle and the Manhattan radius R. Specifically, assuming that p i and its adjacent sub-image matrix p i ' come from the same distribution, the set of adjacent sub-images Z i can include: ( 2R+1) 2 adjacent sub-image matrices p i '.
  • Step c determine the distribution function f i (p i ) of p i by formula one, formula one may include:
  • h represents the bandwidth between p i and p i '.
  • calculating the self-information of each sub-image matrix based on the distribution function may include:
  • Formula 2 to calculate the self-information I i (p i ) of the i-th sub-image matrix based on the distribution function f i (p i ) of the i-th sub-image matrix.
  • Formula 2 includes:
  • the self-information amount corresponding to each sub-image matrix can be determined through the above method, and then a self-information amount matrix can be formed based on the self-information amount of each sub-image matrix.
  • the method of composing a self-information matrix based on the self-information of each sub-image matrix may include
  • Step 1 Establish a first empty matrix whose size is the same as that of the CSI image information matrix.
  • Step 2 According to the position of each sub-image matrix in the extended CSI image information matrix, fill the self-information corresponding to each sub-image matrix into the first empty matrix to form a self-information matrix.
  • the extended CSI image information matrix is divided into m sub-image matrices, where m is consistent with the number of elements included in the CSI image information matrix.
  • the number of self-information quantities determined in step 302 is also consistent with the number of elements included in the CSI image information matrix.
  • the size of the first empty matrix is the same as the size of the CSI image information matrix, the number of self-information amounts and the number of empty positions in the first empty matrix should be in one-to-one correspondence.
  • the self-information amount corresponding to each sub-image matrix can be filled into the first empty matrix in a one-to-one correspondence to form a self-information amount matrix.
  • the method of filling the self-information corresponding to each sub-image matrix into the first empty matrix according to the position of each sub-image matrix in the extended CSI image information matrix may include:
  • the self-information corresponding to the sub-image matrix can be filled into row W and column Q of the first empty matrix.
  • Step 204 Replace the elements in the self-information matrix whose element values are smaller than the preset threshold with 0 to obtain a sparse self-information matrix, and determine the position information of non-zero elements in the sparse self-information matrix in the sparse self-information matrix.
  • the preset threshold may be preset.
  • the method for determining elements whose element values are smaller than a preset threshold from the information amount matrix may include:
  • the attenuation coefficient may obey the Boltzmann distribution, and the attenuation function may be, for example:
  • T can be regarded as a soft threshold.
  • the elements with small element values in the self-information matrix are set to 0.
  • the element values in the self-information matrix are all set with equal probability. 0, that is, randomly set to 0.
  • the value of T in the attenuation function should be relatively small.
  • the attenuation coefficient is inversely proportional to the amount of self-information
  • the corresponding attenuation rate is smaller, and then retained, when the amount of self-information is not greater than the preset
  • the threshold is set, if the attenuation rate is relatively large, it is set to 0, so as to obtain a sparse self-information matrix.
  • the size of the sparse self-information matrix is the same as the size of the CSI image information matrix.
  • the redundant information in the self-information matrix can be removed, so that subsequent information based on the self-information matrix
  • the CSI image matrix is compressed and quantized, it can be easily compressed and the compression efficiency can be ensured.
  • the position information of the non-zero elements in the sparse self-information matrix in the sparse self-information matrix can also be determined, so that the subsequent The UE can send the location information to the base station, so that the base station can reconstruct the CSI information matrix according to the location information, ensuring the accuracy of the reconstructed CSI information matrix.
  • Step 205 using the third convolutional layer to perform feature processing on the sparse self-information matrix to obtain n second feature maps.
  • the third convolutional layer may be a convolutional layer without gradient update.
  • the n second feature maps may be n-dimensional feature maps corresponding to the self-information matrix.
  • n may be equal to 64, that is, the third convolutional layer is used to perform feature processing on the self-information matrix to obtain a 64-dimensional second feature map corresponding to the self-information matrix.
  • the structural principles of the first convolutional layer and the third convolutional layer can be the same, and both are used to perform feature analysis on the input matrix to obtain n corresponding to the input matrix feature map.
  • Step 206 Use the fourth convolutional layer to determine a sparse CSI information matrix based on n first feature maps and n second feature maps, and position information of non-zero elements in the sparse CSI information matrix in the sparse CSI information matrix .
  • the fourth convolutional layer may be a convolutional layer with gradient update.
  • the size of the convolution kernel of the fourth convolution layer may be c ⁇ n ⁇ s ⁇ s, and the convolution step size may be L.
  • the size of the convolution kernel of the fourth convolution layer may be 2 ⁇ 64 ⁇ 3 ⁇ 3, and the convolution step may be 1.
  • the sparse CSI information matrix and the CSI information matrix have the same size and the same dimension.
  • the elements in the CSI information matrix can be screened based on the self-information amount of the CSI information matrix to obtain a sparse CSI information matrix.
  • the above-mentioned scheme may also include:
  • Step 103 Determine feedback information based on elements in the sparse CSI information matrix, and send the feedback information to the base station.
  • Fig. 4 is a schematic flowchart of determining feedback information provided by an embodiment of the present disclosure. As shown in Fig. 4, the method may include:
  • Step 401 Compress and quantize the elements in the sparse CSI information matrix to obtain compressed codewords.
  • the method for compressing and quantizing the elements in the sparse CSI information matrix may include:
  • Step A Sorting the values of each element in the sparse CSI information matrix from large to small according to the amount of self-information.
  • Step B Select the first M element values from the sorted element values to compress and quantize to form a compressed codeword, wherein one element value corresponds to one codeword.
  • the method for determining M may include:
  • Step B1 determining the compression ratio ⁇ .
  • the compression ratio ⁇ may be determined based on user requirements. Exemplarily, in an embodiment of the present disclosure, it may be determined that the compression rate ⁇ is equal to 1/4.
  • Step B2 calculate the code number M of the compressed codeword based on formula three, and said formula three includes:
  • a indicates the number of bits required to transmit a codeword
  • k indicates the number of bits required to transmit the position information of the corresponding element of a codeword
  • N c indicates the number of antennas set by the base station
  • N t indicates the number of antennas corresponding to each antenna number of subcarriers.
  • formula 3 is the same as that in the prior art
  • the formula of the compression rate ⁇ is different, and formula 3 also includes the bit number k required to transmit the position information of a codeword.
  • the classical Lloyd algorithm may be used in step B to compress and quantize the first M element values, and the M element values may be quantized by 8 bits.
  • Step C Determine the position information of each element value in the sparse CSI information matrix among the first M element values as the position information of the compressed element value in the sparse CSI information matrix.
  • Step 402 Determine the position information of the compressed element in the sparse CSI information matrix, and determine the compressed codeword and position information as feedback information.
  • the UE will filter the elements in the CSI information matrix to obtain the sparse CSI information matrix. Specifically, when determining the sparse CSI information matrix, it will first The CSI image information matrix corresponding to the information matrix is divided into multiple sub-image information matrices, and then the distribution function corresponding to each sub-image information matrix is determined according to each sub-image information matrix and adjacent sub-image information matrices, and then determined based on the distribution function The self-information corresponding to each sub-image information matrix is obtained, and finally, the redundant information is screened based on the self-information to determine the sparse CSI information matrix. And, further, the UE determines feedback information according to the determined elements in the sparse CSI information matrix, and sends the feedback information to the base station, so that the base station reconstructs the CSI information matrix based on the feedback information.
  • the sparse CSI information matrix in the embodiment of the present disclosure is a matrix after redundant information is deleted, the convenience of subsequent operations on the sparse CSI information matrix can be ensured, and the overhead can be reduced.
  • the feedback information determined by the UE includes the position information of the compressed elements in the sparse CSI information matrix, so that the subsequent base station can accurately reconstruct the
  • the CSI information matrix further ensures the reconstruction accuracy of the CSI information matrix.
  • FIG. 5 is a schematic flowchart of an information feedback method provided by an embodiment of the present disclosure, which is applied to a base station. As shown in FIG. 5, the information feedback method may include the following steps:
  • Step 501 Obtain feedback information sent by the UE, and determine a preliminary CSI information matrix based on the feedback information.
  • the feedback information may include: compressed codewords obtained by compressing and quantizing the elements in the sparse CSI information matrix corresponding to the CSI information matrix, and the compressed elements in the sparse CSI information matrix Position information in the CSI information matrix.
  • determining the preliminary CSI information matrix based on the feedback information may include:
  • Step (1) dequantize the compressed codeword to obtain the dequantized codeword.
  • the element value corresponding to each codeword can be determined by dequantizing the compressed codeword.
  • Step (2) constructing a second empty matrix, the size of the second empty matrix is the same as that of the CSI information matrix.
  • Step (3) filling the dequantized codewords into the second empty matrix based on the position information.
  • filling the dequantized codewords into the second empty matrix based on the position information may include:
  • the dequantized codeword may be filled into row W and column Q of the second empty matrix.
  • Step (4) calculating the mean value of the dequantized codewords, and filling the mean value into other positions of the second empty matrix to obtain a preliminary CSI information matrix.
  • the dequantized codewords should also only have M elements. Therefore, in one embodiment of the present disclosure, there may be a situation that the number of codewords after inverse quantization is less than the number of positions in the second empty matrix. That is, the dequantized codeword cannot fill the second empty matrix.
  • the mean value of the dequantized codeword may be filled in other positions of the second empty matrix, so as to fill the second empty matrix to obtain a preliminary CSI information matrix.
  • Step 502. Reconstruct the CSI information matrix based on the prepared CSI information matrix.
  • determining the CSI information matrix based on the prepared CSI information matrix may include: inputting the prepared CSI information matrix into a pre-trained convolution structure to output a reconstructed CSI information matrix.
  • the convolutional structure may include sequentially connected fifth convolutional layers, sixth convolutional layers, seventh convolutional layers, eighth convolutional layers, and fourth convolutional layers. Nine convolutional layers, and tenth convolutional layers;
  • the output end of the fifth convolutional layer is also connected to the input end of the eighth convolutional layer, and the output end of the eighth convolutional layer is also connected to the input end of the tenth convolutional layer.
  • the convolution kernel of the fifth layer may be R 1 ⁇ c ⁇ s ⁇ s
  • the convolution kernel of the sixth layer may be R 2 ⁇ R 1 ⁇ s ⁇ s
  • the convolution kernel of the sixth layer may be R 2 ⁇ R 1 ⁇ s ⁇ s.
  • the convolution kernel of the seventh layer can be c ⁇ R 2 ⁇ s ⁇ s
  • the convolution kernel of the eighth layer can be R 1 ⁇ c ⁇ s ⁇ s
  • the convolution kernel of the ninth layer can be R 2 ⁇ R 1 ⁇ s ⁇ s
  • the convolution kernel of the ninth layer can be R 2 ⁇ R 1 ⁇ s ⁇ s.
  • the ten-layer convolution kernel is c ⁇ R 2 ⁇ s ⁇ s, and the fifth convolutional layer, the sixth convolutional layer, the seventh convolutional layer, the eighth convolutional layer, the ninth convolutional layer, and the tenth convolutional layer
  • the strides of the convolutional layers can be the same.
  • the convolution kernel of the fourth layer can be 8 ⁇ 2 ⁇ 3 ⁇ 3
  • the convolution kernel of the fifth layer can be 16 ⁇ 8 ⁇ 3 ⁇ 3
  • the convolution kernel of the sixth layer can be
  • the product kernel can be 2 ⁇ 16 ⁇ 3 ⁇ 3
  • the seventh layer convolution kernel is 8 ⁇ 2 ⁇ 3 ⁇ 3
  • the eighth layer convolution kernel is 16 ⁇ 8 ⁇ 3 ⁇ 3
  • the ninth layer convolution kernel is 2 ⁇ 16 ⁇ 3 ⁇ 3
  • the step size of the fifth convolutional layer, the sixth convolutional layer, the seventh convolutional layer, the eighth convolutional layer, the ninth convolutional layer, and the tenth convolutional layer can be equal to is 1.
  • the convolution operation of the convolution structure can be defined as:
  • y d, i, j is the (d, i, j)th element of the convolutional structure output
  • d is the number of output channels
  • i is the number of rows of y d, i, j in the reconstructed CSI information matrix
  • j is y d
  • i, j are the number of columns in the reconstructed CSI information matrix
  • W d, c, h, w are the (d, c, h, w)th elements in the convolution kernel weight matrix W
  • c is the number of input channels
  • h is the length of the convolution kernel
  • w is the width of the convolution kernel
  • b d is the dth element in the convolution kernel offset b
  • i is The number of rows in the preliminary CSI information matrix
  • j is The number of columns in the preliminary CSI information matrix
  • the reconstructed CSI information matrix specifically includes a reconstructed CSI information matrix corresponding to the real part of the CSI information matrix, and a reconstructed CSI information matrix corresponding to the imaginary part of the CSI information matrix.
  • the reconstructed CSI information matrix corresponding to the real part of the CSI information matrix and the reconstructed CSI information matrix corresponding to the imaginary part of the CSI information matrix is combined into a complex matrix, and then a two-dimensional inverse discrete Fourier transform is performed on the complex matrix to obtain a space-frequency domain CSI information matrix.
  • the UE will filter the elements in the CSI information matrix to obtain the sparse CSI information matrix. Specifically, when determining the sparse CSI information matrix, it will first The CSI image information matrix corresponding to the information matrix is divided into multiple sub-image information matrices, and then the distribution function corresponding to each sub-image information matrix is determined according to each sub-image information matrix and adjacent sub-image information matrices, and then determined based on the distribution function The self-information corresponding to each sub-image information matrix is obtained, and finally, the redundant information is screened based on the self-information to determine the sparse CSI information matrix. And, further, the UE determines feedback information according to the determined elements in the sparse CSI information matrix, and sends the feedback information to the base station, so that the base station reconstructs the CSI information matrix based on the feedback information.
  • the sparse CSI information matrix in the embodiments of the present disclosure is a matrix after redundant information is deleted, the convenience of subsequent operations on the sparse CSI information matrix can be ensured, and overhead can be reduced.
  • the feedback information determined by the UE includes the position information of the compressed elements in the sparse CSI information matrix, so that the subsequent base station can accurately reconstruct the
  • the CSI information matrix further ensures the reconstruction accuracy of the CSI information matrix.
  • FIG. 6 is a schematic structural diagram of an information feedback device provided by an embodiment of the present disclosure, which is configured in a UE, as shown in FIG. 6 , and may include:
  • the first acquiring module is configured to acquire the channel state information CSI information matrix.
  • a screening module configured to screen elements in the CSI information matrix based on the amount of self-information of the CSI information matrix to obtain a sparse CSI information matrix
  • a feature encoder configured to determine feedback information based on elements in the sparse CSI information matrix, and send the feedback information to the base station.
  • the UE will filter the elements in the CSI information matrix to obtain the sparse CSI information matrix. Specifically, when determining the sparse CSI information matrix, it will first The CSI image information matrix corresponding to the information matrix is divided into multiple sub-image information matrices, and then the distribution function corresponding to each sub-image information matrix is determined according to each sub-image information matrix and adjacent sub-image information matrices, and then determined based on the distribution function The self-information corresponding to each sub-image information matrix is obtained, and finally, the redundant information is screened based on the self-information to determine the sparse CSI information matrix. And, further, the UE determines feedback information according to the determined elements in the sparse CSI information matrix, and sends the feedback information to the base station, so that the base station reconstructs the CSI information matrix based on the feedback information.
  • the sparse CSI information matrix in the embodiments of the present disclosure is a matrix after redundant information is deleted, the convenience of subsequent operations on the sparse CSI information matrix can be ensured, and overhead can be reduced.
  • the feedback information determined by the UE includes the position information of the compressed elements in the sparse CSI information matrix, so that the subsequent base station can accurately reconstruct the
  • the CSI information matrix further ensures the reconstruction accuracy of the CSI information matrix.
  • the first acquiring module is further configured to: transform the acquired CSI information matrix from the space-frequency domain to the angular delay domain, wherein the CSI information A matrix includes real and imaginary parts.
  • the screening module is also used for:
  • n is an integer and n ⁇ 1;
  • the size of the self-information matrix is the same as the size of the CSI image information matrix
  • the fourth convolutional layer to determine a sparse CSI information matrix based on the n first feature maps and the n second feature maps, and the non-zero elements in the sparse CSI information matrix are included in the sparse CSI information matrix location information.
  • the screening module is also used for:
  • the screening module is also used for:
  • the adjacent sub-image set Z i includes: centered on the p i , centered on Manhattan (2R+1) 2 adjacent sub-image matrices p i ' within a circle whose radius R is the radius;
  • the distribution function f i (p i ) of the p i is determined by formula one, and the formula one includes:
  • h represents the bandwidth between p i and p i '.
  • the screening module is also used for:
  • formula two Utilize formula two to calculate the self-information I i (p i ) of the ith sub-image matrix based on the distribution function f i (p i ) of the i-th sub-image matrix, the formula two includes:
  • the screening module is also used for:
  • the size of the first empty matrix is the same as the size of the CSI image information matrix
  • the self-information amount corresponding to each sub-image matrix is filled into the first empty matrix to form a self-information amount matrix.
  • both the first convolutional layer and the fourth convolutional layer are convolutional layers with gradient update.
  • the device is further configured to: train the first convolutional layer and the fourth convolutional layer.
  • the feature encoder further includes:
  • a compression quantization encoder configured to compress and quantize elements in the sparse CSI information matrix to obtain compressed codewords
  • the determining unit is further configured to determine position information of compressed elements in the sparse CSI information matrix, and determine the compressed codeword and the position information as the feedback information.
  • the device further includes a first dimension mapping module connected between the screening module and the feature encoder, for when the dimension of the sparse CSI information matrix output by the screening module is not When applicable to the dimension of the feature encoder, the dimension of the sparse CSI information matrix output by the screening module is mapped to the dimension applicable to the feature encoder.
  • FIG. 7 is a schematic structural diagram of an information feedback device provided by an embodiment of the present disclosure, which is configured in a base station. As shown in FIG. 7 , it may include:
  • a second acquiring module configured to acquire feedback information sent by the UE, and determine a preliminary CSI information matrix based on the feedback information
  • a reconstruction module configured to reconstruct a CSI information matrix based on the prepared CSI information matrix.
  • the UE will filter the elements in the CSI information matrix to obtain the sparse CSI information matrix. Specifically, when determining the sparse CSI information matrix, it will first The CSI image information matrix corresponding to the information matrix is divided into multiple sub-image information matrices, and then the distribution function corresponding to each sub-image information matrix is determined according to each sub-image information matrix and adjacent sub-image information matrices, and then determined based on the distribution function The self-information corresponding to each sub-image information matrix is obtained, and finally, the redundant information is screened based on the self-information to determine the sparse CSI information matrix. And, further, the UE determines feedback information according to the determined elements in the sparse CSI information matrix, and sends the feedback information to the base station, so that the base station reconstructs the CSI information matrix based on the feedback information.
  • the sparse CSI information matrix in the embodiments of the present disclosure is a matrix after redundant information is deleted, the convenience of subsequent operations on the sparse CSI information matrix can be ensured, and overhead can be reduced.
  • the feedback information determined by the UE includes the position information of the compressed elements in the sparse CSI information matrix, so that the subsequent base station can accurately reconstruct the
  • the CSI information matrix further ensures the reconstruction accuracy of the CSI information matrix.
  • the feedback information includes: compressed codewords obtained by compressing and quantizing elements in the sparse CSI information matrix corresponding to the CSI information matrix, and compressed elements Position information in the sparse CSI information matrix;
  • the second acquisition module also includes:
  • An inverse quantizer configured to inverse quantize the compressed codeword to obtain an inverse quantized codeword
  • An empty matrix construction unit configured to construct a second empty matrix, the size of the second empty matrix is the same as the size of the CSI information matrix;
  • an interpolation unit configured to fill the dequantized codeword into the second empty matrix based on the position information
  • the mean value filling unit is configured to calculate the mean value of the dequantized codewords, and fill the mean value into other positions of the second empty matrix to obtain a preliminary CSI information matrix.
  • the reconstruction module is also used for:
  • the convolutional structure includes a fifth convolutional layer, a sixth convolutional layer, a seventh convolutional layer, an eighth convolutional layer, and a ninth convolutional layer connected in sequence stack layer, and the tenth convolutional layer;
  • the output end of the fifth convolutional layer is also connected to the input end of the eighth convolutional layer, and the output end of the eighth convolutional layer is also connected to the input end of the tenth convolutional layer.
  • the device is further configured to: train the convolutional structure.
  • the reconstruction module further includes a second dimension mapping module connected to the output end of the tenth convolutional layer, for mapping the dimension of the matrix output by the tenth convolutional layer is the dimension of the original CSI information matrix.
  • the reconstructed CSI information matrix includes a reconstructed CSI information matrix corresponding to the real part of the CSI information matrix and a reconstructed CSI information matrix corresponding to the imaginary part of the CSI information matrix;
  • the reconstruction module is further configured to combine the reconstructed CSI information matrix corresponding to the real part of the CSI information matrix and the reconstructed CSI information matrix corresponding to the imaginary part of the CSI information matrix into a complex matrix, and then binary the complex matrix
  • the space-frequency domain CSI information matrix is obtained by inverse discrete Fourier transform.
  • FIG. 8 is a schematic structural diagram of an information feedback system model provided by an embodiment of the present disclosure. As shown in FIG. 8 , it includes a first acquisition module, a screening module, a first dimension mapping module, a feature encoder, a second Two acquisition modules and reconstruction modules, wherein, the detailed introduction of the first acquisition module, screening module, second dimension mapping module, feature encoder, second acquisition module, and reconstruction module can refer to the above description, and the embodiments of the present disclosure are here I won't go into details.
  • the UE will filter the elements in the CSI information matrix to obtain the sparse CSI information matrix. Specifically, when determining the sparse CSI information matrix, it will first The CSI image information matrix corresponding to the CSI information matrix is divided into multiple sub-image information matrices, and then the distribution function corresponding to each sub-image information matrix is determined according to each sub-image information matrix and adjacent sub-image information matrices, and then based on the distribution function The self-information corresponding to each sub-image information matrix is determined, and finally, the redundant information is screened based on the self-information to determine a sparse CSI information matrix. And, further, the UE determines feedback information according to the determined elements in the sparse CSI information matrix, and sends the feedback information to the base station, so that the base station reconstructs the CSI information matrix based on the feedback information.
  • the sparse CSI information matrix in the embodiments of the present disclosure is a matrix after redundant information is deleted, the convenience of subsequent operations on the sparse CSI information matrix can be ensured, and overhead can be reduced.
  • the feedback information determined by the UE includes the position information of the compressed elements in the sparse CSI information matrix, so that the subsequent base station can accurately reconstruct the
  • the CSI information matrix further ensures the reconstruction accuracy of the CSI information matrix.
  • the information feedback system model shown in FIG. 8 may be trained, wherein the training method may include:
  • Step 1 first obtain a CSI information matrix sample set, wherein the CSI information matrix sample set may include training samples, verification samples, and test samples.
  • the COST2100 [7] channel model can be used to generate 150,000 space-frequency domain CSI information matrix samples in a 5.3GHz indoor micro-cell scenario.
  • the training samples may include 100,000
  • the verification samples may include 30,000
  • the test samples may include 20,000.
  • Step 2 Using the methods shown in the above-mentioned Figures 1 to 5 to train the information feedback system model based on the CSI information matrix sample set, and calculate the loss function L.
  • the loss function L can be defined as:
  • N is the number of training samples
  • H a is the original CSI information matrix acquired by the first acquisition module configured in the UE
  • is the Euclidean norm.
  • Step 3 Update the parameters of the information feedback system model according to the loss function.
  • the parameters of the information feedback system model may include weights and offsets of each convolutional layer.
  • the embodiment of the present disclosure adopts the methods shown in FIGS. 1 to 5 to train the information feedback system model, wherein the redundant information of the matrix based on the methods shown in FIGS. 1 to 5 is relatively Less, lower overhead, so that the complexity of the training method for the information feedback system model is also lower, and the convergence is faster, which improves the accuracy and efficiency of the training method.
  • the output of the fifth convolutional layer in the reconstruction module is also connected to the input of the eighth convolutional layer, and the output of the eighth convolutional layer is also connected to the input of the tenth convolutional layer , it can prevent the phenomenon of gradient disappearance during the training of the reconstruction module and ensure the training accuracy.
  • the computer storage medium provided by the embodiments of the present disclosure stores an executable program; after the executable program is executed by a processor, the method shown in any one of FIG. 1 to FIG. 4 or FIG. 5 can be implemented.
  • the present disclosure further proposes a computer program product, including a computer program, and when the computer program is executed by a processor, the method shown in any one of FIG. 1 to FIG. 4 or FIG. 5 is implemented.
  • the present disclosure further proposes a computer program.
  • the program When the program is executed by a processor, the method as shown in any one of FIG. 1 to FIG. 4 or FIG. 5 can be realized.
  • Fig. 9 is a block diagram of a user equipment UE900 provided by an embodiment of the present disclosure.
  • the UE 900 may be a mobile phone, a computer, a digital broadcast terminal device, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, and the like.
  • UE900 may include at least one of the following components: a processing component 902, a memory 904, a power supply component 906, a multimedia component 908, an audio component 910, an input/output (I/O) interface 912, a sensor component 913, and a communication component 916.
  • a processing component 902 a memory 904
  • a power supply component 906 a multimedia component 908, an audio component 910, an input/output (I/O) interface 912, a sensor component 913, and a communication component 916.
  • I/O input/output
  • the processing component 902 generally controls the overall operations of the UE 900, such as those associated with display, phone calls, data communications, camera operations, and recording operations.
  • the processing component 902 may include at least one processor 920 to execute instructions to complete all or part of the steps of the above-mentioned method.
  • processing component 902 can include at least one module to facilitate interaction between processing component 902 and other components.
  • processing component 902 may include a multimedia module to facilitate interaction between multimedia component 908 and processing component 902 .
  • the memory 904 is configured to store various types of data to support operations at the UE 900 . Examples of such data include instructions for any application or method operating on UE900, contact data, phonebook data, messages, pictures, videos, etc.
  • the memory 904 can be implemented by any type of volatile or non-volatile memory device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read-only memory
  • EPROM erasable Programmable Read Only Memory
  • PROM Programmable Read Only Memory
  • ROM Read Only Memory
  • Magnetic Memory Flash Memory
  • Magnetic or Optical Disk Magnetic Disk
  • the power supply component 906 provides power to various components of the UE 900 .
  • Power component 906 may include a power management system, at least one power supply, and other components associated with generating, managing, and distributing power for UE 900 .
  • the multimedia component 908 includes a screen providing an output interface between the UE 900 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user.
  • the touch panel includes at least one touch sensor to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense a boundary of a touch or slide action, but also detect a wake-up time and pressure related to the touch or slide operation.
  • the multimedia component 908 includes a front camera and/or a rear camera. When the UE900 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capability.
  • the audio component 910 is configured to output and/or input audio signals.
  • the audio component 910 includes a microphone (MIC), which is configured to receive an external audio signal when the UE 900 is in an operation mode, such as a call mode, a recording mode, and a voice recognition mode. Received audio signals may be further stored in memory 904 or sent via communication component 916 .
  • the audio component 910 also includes a speaker for outputting audio signals.
  • the I/O interface 912 provides an interface between the processing component 902 and a peripheral interface module.
  • the peripheral interface module may be a keyboard, a click wheel, a button, and the like. These buttons may include, but are not limited to: a home button, volume buttons, start button, and lock button.
  • the sensor component 913 includes at least one sensor for providing various aspects of state assessment for the UE 900 .
  • the sensor component 913 can detect the open/closed state of the device 900, the relative positioning of components, such as the display and keypad of the UE900, the sensor component 913 can also detect the position change of the UE900 or a component of the UE900, and the user and Presence or absence of UE900 contact, UE900 orientation or acceleration/deceleration and temperature change of UE900.
  • the sensor assembly 913 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact.
  • the sensor assembly 913 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor component 913 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor or a temperature sensor.
  • Communication component 916 is configured to facilitate wired or wireless communications between UE 900 and other devices.
  • UE900 can access wireless networks based on communication standards, such as WiFi, 2G or 3G, or a combination thereof.
  • the communication component 916 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communication component 916 also includes a near field communication (NFC) module to facilitate short-range communication.
  • NFC near field communication
  • the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (BT) technology and other technologies.
  • RFID Radio Frequency Identification
  • IrDA Infrared Data Association
  • UWB Ultra Wide Band
  • Bluetooth Bluetooth
  • UE 900 may be powered by at least one Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array ( FPGA), controller, microcontroller, microprocessor or other electronic components for implementing the above method.
  • ASIC Application Specific Integrated Circuit
  • DSP Digital Signal Processor
  • DSPD Digital Signal Processing Device
  • PLD Programmable Logic Device
  • FPGA Field Programmable Gate Array
  • controller microcontroller, microprocessor or other electronic components for implementing the above method.
  • Fig. 10 is a block diagram of a base station 1000 provided by an embodiment of the present disclosure.
  • base station 1000 may be provided as a base station.
  • base station 1000 includes processing component 1026 , which further includes at least one processor, and a memory resource represented by memory 1032 for storing instructions executable by processing component 1022 , such as application programs.
  • the application program stored in memory 1032 may include one or more modules each corresponding to a set of instructions.
  • the processing component 1026 is configured to execute instructions, so as to execute any of the aforementioned methods applied to the base station, for example, the method shown in FIG. 1 .
  • Base station 1000 may also include a power component 1026 configured to perform power management of base station 1000, a wired or wireless network interface 1050 configured to connect base station 1000 to a network, and an input-output (I/O) interface 1058.
  • the base station 1000 can operate based on an operating system stored in the memory 1032, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, Free BSDTM or similar.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本公开提出一种信息反馈方法、装置、用户设备、基站、系统模型及存储介质,其中,方法包括:获取信道状态信息CSI信息矩阵;基于所述CSI信息矩阵的自信息量对所述CSI信息矩阵中的元素进行筛选,得到稀疏CSI信息矩阵;基于所述稀疏CSI信息矩阵中的元素确定反馈信息,并将所述反馈信息发送至基站。本公开提出的方法中基站重建矩阵的准确度较高且开销较低。

Description

信息反馈方法、装置、用户设备、基站、系统模型及存储介质 技术领域
本公开涉及通信技术领域,尤其涉及一种信息反馈方法、装置、用户设备、基站、系统模型及存储介质。
背景技术
由于m-MIMO(massive Multiple-Input Multiple-Output,大规模多输入多输出)技术的稳定性、能量利用率、以及抗干扰能力均较好,因此,通常会利用m-MIMO系统进行无线通信。其中,在m-MIMO系统中,UE(User Equipment,用户设备)通常需要向基站反馈下行链路的CSI(ChannelState Information,信道状态信息)信息矩阵,以使得基站可以根据该CSI信息矩阵确定下行链路的信道质量。其中,由于m-MIMO系统中基站端天线数量较多,则导致基站端对应的下行链路的数量也较多,从而导致反馈CSI信息矩阵的开销较大,因此,亟需一种低开销信息反馈方法。
相关技术中,UE反馈CSI信息矩阵的方法主要包括:
方法一:利用AoD(Angle-of-departure,偏移角)自适应子空间码本将CSI信息矩阵量化成一定数量的比特数,并反馈至基站,使得基站基于该一定数量的比特数重建CSI信息矩阵。
方法二:将CSI信息矩阵变换至信道空间下的稀疏矩阵,并对该稀疏矩阵进行随机压缩采样以获得低维度测量值,并反馈给基站,使得基站基于该低维度测量值重建CSI信息矩阵。
方法三:基于DL(Deep Learning,深度学习)的神经网络模型向基站反馈CSI信息矩阵。
其中,方法一中当CSI信息矩阵所需转换的比特数较多时,其开销仍然较大。方法二中在获得低纬度测量值时是随机采样的,未考虑到CSI信息矩阵的结构特性,使得所获得的低纬度测量值无法准确反映出原始CSI信息矩阵,从而使得基站重建的CSI信息矩阵与UE发送的原始CSI信息矩阵存在较大差异,导致基站重建准确度较低。方法三中训练神经网络模型时的复杂度较高,且收敛慢,同时,基站重建准确度也较低。
发明内容
本公开提出的信息反馈方法、装置、用户设备、基站、系统模型及存储介质,以解决现有的信息反馈方法中基站重建准确度较低以及开销较大的技术问题。
本公开第一方面实施例提出的信息反馈方法,应用于UE,包括:
获取信道状态信息CSI信息矩阵;
基于所述CSI信息矩阵的自信息量对所述CSI信息矩阵中的元素进行筛选,得到稀疏CSI信息矩阵;
基于所述稀疏CSI信息矩阵中的元素确定反馈信息,并将所述反馈信息发送至基站。
本公开第二方面实施例提出的信息反馈方法,应用于基站,包括:
获取UE发送的反馈信息,并基于所述反馈信息确定预备CSI信息矩阵;
基于所述预备CSI信息矩阵重建CSI信息矩阵。
本公开第三方面实施例提出的信息反馈装置,包括:
第一获取模块,用于获取信道状态信息CSI信息矩阵;
筛选模块,用于基于所述CSI信息矩阵的自信息量对所述CSI信息矩阵中的元素进行筛选,得到稀疏CSI信息矩阵;
特征编码器,用于基于所述稀疏CSI信息矩阵中的元素确定反馈信息,并将所述反馈信息发送至基站。
本公开第四方面实施例提出的信息反馈装置,包括:
第二获取模块,用于获取UE发送的反馈信息,并基于所述反馈信息确定预备CSI信息矩阵;
重建模块,用于基于所述预备CSI信息矩阵重建CSI信息矩阵。
本公开第五方面实施例提出的一种用户设备,包括:收发器;存储器;处理器,分别与所述收发器及所述存储器连接,配置为通过执行所述存储器上的计算机可执行指令,控制所述收发器的无线信号收发,并能够实现如上第一方面实施例提出的方法。
本公开第六方面实施例提出的一种基站,其中,包括:收发器;存储器;处理器,分别与所述收发器及所述存储器连接,配置为通过执行所述存储器上的计算机可执行指令,控制所述收发器的无线信号收发,并能够实现如上第二方面实施例提出的方法。
本公开第七方面实施例提出的一种信息反馈系统模型,所述系统模型至少包括如上第三方面和第四方面所述的信息反馈装置。
本公开第八方面实施例提出的一种用于对第七方面所述的信息反馈系统模型的训练方法,包括:通过执行如上第一方面和/或第二方面所述的方法以对所述信息反馈系统模型进行训练。
本公开又一方面实施例提出的计算机存储介质,其中,所述计算机存储介质存储有计算机可执行指令;所述计算机可执行指令被处理器执行后,能够实现如上所述的方法。
综上所述,在本公开实施例提供的信息反馈方法之中,UE会对CSI信息矩阵中的元素进行筛选得到稀疏CSI信息矩阵,具体的,在确定稀疏CSI信息矩阵时,会先对CSI信息矩阵对应的CSI图像信息矩阵进行图像划分得到多个子图像信息矩阵,之后,会根据每个子图像信息矩阵和邻近子图像信息矩阵确定出每个子图像信息矩阵对应的分布函数,再基于分布函数确定出每个子图像信息矩阵对应的自信息量,最后,基于自信息量对冗余信息进行筛选以确定出稀疏CSI信息矩阵。以及,进一步地,UE会根据确定出的稀疏CSI信息矩阵中的元素确定反馈信息,并将反馈信息发送至基站,以使得基站基于该反馈信息重建CSI信息矩阵。
由此可知,本公开实施例中,在确定分布函数时,会考虑到每个子图像矩阵与邻近子图像矩阵之间的结构相关性,则当后续基于分布函数确定稀疏CSI信息矩阵时,不会破坏原始CSI信息矩阵的结构性,从而后续基站利用根据稀疏CSI信息矩阵确定出的反馈信息重建CSI信息矩阵时,可以确保重建的CSI信息矩阵的准确度。
同时,由于本公开实施例中的稀疏CSI信息矩阵是删除了冗余信息之后的矩阵,则可以确保后续对于稀疏CSI信息矩阵操作的便捷性,且降低了开销。
此外,本公开实施例中UE所确定出的反馈信息包括有被压缩的元素在稀疏CSI信息矩阵中的位置信息,从而后续基站可以根据被压缩的元素在稀疏CSI信息矩阵中的位置信息准确重建CSI信息矩阵,则进一步确保了CSI信息矩阵的重建准确度。
本公开附加的方面和优点将在下面的描述中部分给出,部分将从下面的描述中变得明显,或通过本公开的实践了解到。
附图说明
本公开上述的和/或附加的方面和优点从下面结合附图对实施例的描述中将变得明显和容易理解,其中:
图1为本公开实施例所提供的一种信息反馈方法的流程示意图;
图2为本公开一个实施例所提供的一种得到稀疏CSI信息矩阵的流程示意图;
图3为本公开一个实施例所提供的一种利用第二卷积层确定CSI图像信息矩阵对应的自信息量矩阵的流程示意图;
图4为本公开一个实施例所提供的确定反馈信息的流程示意图;
图5为本公开实施例所提供的一种信息反馈方法的流程示意图;
图6为本公开实施例提供的一种信息反馈装置的结构示意图;
图7为本公开实施例提供的一种信息反馈装置的结构示意图;
图8为本公开实施例提供的一种信息反馈系统模型的结构示意图;
图9是本公开一个实施例所提供的一种用户设备的框图;
图10为本公开一个实施例所提供的一种基站的框图。
具体实施方式
这里将详细地对示例性实施例进行说明,其示例表示在附图中。下面的描述涉及附图时,除非另有表示,不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本公开实施例相一致的所有实施方式。相反,它们仅是与如所附权利要求书中所详述的、本公开实施例的一些方面相一致的装置和方法的例子。
在本公开实施例使用的术语是仅仅出于描述特定实施例的目的,而非旨在限制本公开实施例。在本公开实施例和所附权利要求书中所使用的单数形式的“一种”和“该”也旨在包括多数形式,除非上下文清楚地表示其他含义。还应当理解,本文中使用的术语“和/或”是指并包含一个或多个相关联的列出项目的任何或所有可能组合。
应当理解,尽管在本公开实施例可能采用术语第一、第二、第三等来描述各种信息,但这些信息不应限于这些术语。这些术语仅用来将同一类型的信息彼此区分开。例如,在不脱离本公开实施例范围的情况下,第一信息也可以被称为第二信息,类似地,第二信息也可以被称为第一信息。取决于语境,如在此所使用的词语“如果”及“若”可以被解释成为“在……时”或“当……时”或“响应于确定”。
下面详细描述本公开的实施例,所述实施例的示例在附图中示出,其中自始至终相同或类似的标号表示相同或类似的要素。下面通过参考附图描述的实施例是示例性的,旨在用于解释本公开,而不能理解为对本公开的限制。
下面参考附图对本公开提供的信息反馈方法、装置、用户设备、基站、系统模型及存储介质进行详细描述。
图1为本公开实施例所提供的一种信息反馈方法的流程示意图,应用于UE,如图1所示,该信息反馈方法可以包括以下步骤:
步骤101、获取CSI信息矩阵。
需要说明的是,本公开实施例的指示方法可以应用在任意的UE中。UE可以是指向用户提供语音和/或数据连通性的设备。UE可以经RAN(Radio Access Network,无线接入网)与一个或多个核心网进行通信,UE可以是物联网终端,如传感器设备、移动电话(或称为“蜂窝”电话)和具有物联网终端的计算机,例如,可以是固定式、便携式、袖珍式、手持式、计算机内置的或者车载的装置。例如,站(Station,STA)、订户单元(subscriber unit)、订户站(subscriber station),移动站(mobile station)、移动台(mobile)、远程站(remote station)、接入点、远程终端(remoteterminal)、接入终端(access terminal)、用户装置(user terminal)或用户代理(useragent)。或者,UE也可以是无人飞行器的设备。或者,UE也可以是车载设备,比如,可以是具有无线通信功能的行车电脑,或者是外接行车电脑的无线终端。或者,UE也可以是路边设备,比如,可以是具有无线通信功能的路灯、信号灯或者其它路边设备等。
需要说明的是,在本公开的一个实施例之中,UE端可以配置单天线,基站端可以配置多天线。其中,在本公开的一个实施例之中,基站端的天线可以是按照ULA(Uniform Linear Array,均匀线性阵列)方式排列的,示例的,例如可以在半波长间隔配置N t=32根天线。
基于此,在本公开的一个实施例之中,UE获取CSI信息矩阵可以包括:UE获取基站的每个天线通道对应的CSI信息矩阵。
在本公开的另一个实施例之中,UE获取CSI信息矩阵的方法可以包括:UE根据基站发送的导频信息确定出原始CSI信息矩阵,之后,将所获取到的原始CSI信息矩阵从空频域变换到角度时延域以得到角度时延域的CSI信息矩阵。
其中,在本公开的一个实施例之中,UE可以利用二维离散傅里叶变换将原始CSI信息矩阵
Figure PCTCN2021098711-appb-000001
从空频域变换到角度时延域,其中,角度时延域的CSI信息矩阵
Figure PCTCN2021098711-appb-000002
F d
Figure PCTCN2021098711-appb-000003
Figure PCTCN2021098711-appb-000004
N t×N t对应的离散傅里叶变换矩阵,
Figure PCTCN2021098711-appb-000005
指示基站所设置的天线数,例如m-MIMO系统采用OFDM(Orthogonal Frequency Division Multiplexing,正交频分复用)技术,
Figure PCTCN2021098711-appb-000006
以及,N t指示每根天线数对应的子载波数,例如N t=32,上标H表示矩阵的共轭转置。以及,在本公开的一个实施例之中,角度时延域的 CSI信息矩阵H的大小可以为d 1×N c×N t,其中,d 1为天线通道数,N c指示基站设置的天线数。以及,角度时延域的CSI信息矩阵H可以包括实部矩阵H re和虚部矩阵H im
需要说明的是,在本公开的一个实施例之中,后续对角度时延域的CSI信息矩阵H的处理(例如后续的筛选、压缩量化、填充、以及重建等处理步骤)具体是分别对CSI信息矩阵H的实部矩阵H re和虚部矩阵H im进行处理的。
在一些实施例中,上述的方案还可以包括:
步骤102、基于CSI信息矩阵的自信息量对CSI信息矩阵中的元素进行筛选,得到稀疏CSI信息矩阵。
需要说明的是,步骤101可以单独被实施,也可以与步骤102一起被实施。即:任何设备可以只执行步骤101,以获取CSI信息矩阵。任何设备也可以执行步骤101和步骤102,以获取CSI信息矩阵,并根据该于CSI信息矩阵得到稀疏CSI信息矩阵。
其中,图2为本公开一个实施例所提供的一种得到稀疏CSI信息矩阵的流程示意图,如图2所示,该方法可以包括:
步骤201、将CSI信息矩阵映射为CSI图像信息矩阵。
在本公开的一个实施例中,可以通过任何恰当的方式获取该CSI信息矩阵;即:可以采用如前述的步骤101相同的方式获取CSI信息矩阵,也可以采用其他任何可行的方式获取CSI信息矩阵。
在本公开的一个实施例之中,将CSI信息矩阵映射为CSI图像信息矩阵可以包括:将CSI信息矩阵H的实部Hre转换为实部Hre对应的CSI图像信息矩阵,以及将CSI信息矩阵H的虚部Him转换为虚部Him对应的CSI图像信息矩阵。
步骤202、利用第一卷积层对CSI图像信息矩阵进行特征处理,得到n个第一特征图,n为整数,n≥1。
其中,在本公开的一个实施例之中,该第一卷积层可以为带梯度更新的卷积层。以及,示例的,在本公开的一个实施例之中,该第一卷积层的卷积核大小可以为n×c×s×s,卷积步长可以为L,其中,L<s,n、c、s、L均为正整数。示例的,在本公开的一个实施例之中,第一卷积层的卷积核大小可以为64×2×3×3,卷积步长可以为1。
以及,在本公开的一个实施例之中,该n个第一特征图可以为CSI图像信息矩阵对应的n维特征图。示例的,在本公开的一个实施例之中,当n等于64时,则说明:利用第一卷积层对CSI图像信息矩阵进行特征处理,得到该CSI图像信息矩阵对应的64维第一特征图。
步骤203、利用第二卷积层确定CSI图像信息矩阵对应的自信息量矩阵,自信息量矩阵的大小与CSI信息矩阵的大小相同。
其中,图3为本公开一个实施例所提供的一种利用第二卷积层确定CSI图像信息矩阵对应的自信息量矩阵的流程示意图,如图3所示,该方法可以包括
步骤301、在CSI图像信息矩阵外周添加0,以扩充CSI图像信息矩阵得到扩充CSI图像信息矩阵,将扩充CSI图像信息矩阵划分为m个子图像矩阵;其中,m为整数,m≥2,且m与CSI图像信息矩阵中所包括的元素数量一致。
在本公开的一个实施例中,可以通过任何恰当的方式获取CSI图像信息矩阵;即:可以采用如前述的步骤201相同的方式获取CSI图像信息矩阵,也可以采用其他任何可行的方式获取CSI信息矩阵。
其中,在本公开的一个实施例之中,将扩充CSI图像信息矩阵划分为m个子图像矩阵的方法可以包括:利用x×s网格对该扩充CSI图像信息矩阵进行划分,其中,x、s均为正整数,x和s可以相同,也可以不同。示例的,在本公开的一个实施例之中,可以用7×7网格对该扩充CSI图像信息矩阵进行划分。
步骤302、利用第二卷积层基于每个子图像矩阵以及每个子图像矩阵的邻近子图像矩阵计算出每个子图像矩阵的分布函数,基于分布函数计算出每个子图像矩阵的自信息量,并基于每个子图像矩阵的自信息量组成自信息量矩阵。
其中,在本公开的一个实施例之中,该第二卷积层可以为无梯度更新的卷积层。
以及,在本公开的一个实施例之中,利用第二卷积层基于每个子图像矩阵以及每个子图像矩阵的邻近子图像矩阵计算出每个子图像矩阵的分布函数的方法可以包括以下步骤:
步骤a、选取输入至第二卷积层的第i个子图像矩阵p i,i为整数,i≥1。
步骤b、确定p i的邻近子图像矩阵p i’,得到p i对应的邻近子图像集合Z i
其中,在本公开的一个实施例之中,可以是基于邻近同分布原则和曼哈顿半径R来确定邻近子图像集合Z i。具体的,假设p i与其邻近子图像矩阵p i’来自相同的分布,则邻近子图像集合Z i可以包括:以p i为中心、以曼哈顿半径R为半径的圆内的(2R+1) 2个邻近子图像矩阵p i’。
步骤c、通过公式一确定p i的分布函数f i(p i),公式一可以包括:
Figure PCTCN2021098711-appb-000007
其中,
Figure PCTCN2021098711-appb-000008
h表示p i与p i’之间的带宽。
则由上述步骤a-c可知,本公开实施例在计算每个子图像矩阵的分布函数时,会根据该子图像矩阵及其邻近子图像矩阵计算,从而可以考虑到各个子图像矩阵之间的结构相关性,也即是,考虑到了CSI图像信息矩阵中各个元素之间的结构相关性,从而当后续基于分布函数对CSI图像信息矩阵压缩量化得到压缩码字时,可以确保CSI信息矩阵的结构相关性不被破坏,则当后续基站基于压缩码字准确重建CSI信息矩阵时,确保了重建准确度。
进一步地,在本公开的一个实施例之中,基于分布函数计算出每个子图像矩阵的自信息量,可以包括:
利用公式二基于第i个子图像矩阵的分布函数f i(p i)计算出第i个子图像矩阵的自信息量I i(p i),公式二包括:
I i(p i)=-log(f i(p i))。
则通过上述方法可以确定出每个子图像矩阵对应的自信息量,之后,即可基于每个子图像矩阵的自信息量组成自信息量矩阵。其中,在本公开的一个实施例之中,基于每个子图像矩阵的自信息量组成自信息量矩阵的方法可以包括
步骤1、建立一第一空矩阵,该第一空矩阵的大小与CSI图像信息矩阵的大小相同。
步骤2、按照每个子图像矩阵在扩充CSI图像信息矩阵的位置将每个子图像矩阵对应的自信息量填充至第一空矩阵中,以组成自信息量矩阵。
其中,需要说明的是,在本公开的一个实施例之中,步骤301中将扩充CSI图像信息矩阵划分为了m个子图像矩阵,其中,m与CSI图像信息矩阵中所包括的元素数量一致。以及,由于每个子图像矩阵均对应一个自信息量,因此,步骤302中所确定出自信息量的个数与CSI图像信息矩阵中所包括的元素数量也是一致的。基于此,在第一空矩阵的大小与CSI图像信息矩阵的大小相同的前提下,自信息量的个数与第一空矩阵中的空位置数应当是一一对应的。由此,按照每个子图像矩阵在扩充CSI图像信息矩阵的位置可以将每个子图像矩阵对应的自信息量一一对应填充至第一空矩阵中,组成自信息量矩阵。
示例的,在本公开的一个实施例之中,按照每个子图像矩阵在扩充CSI图像信息矩阵的位置将每个子图像矩阵对应的自信息量填充至所述第一空矩阵中的方法可以包括:
若某一子图像矩阵在扩充CSI图像信息矩阵位于第W行、第Q列,则可将该子图像矩阵对应的自信息量填充至第一空矩阵的第W行、第Q列处。
步骤204、将自信息量矩阵中元素值小于预设阈值的元素用0代替得到稀疏自信息量矩阵,并确定稀疏自信息量矩阵中的非0元素在稀疏自信息量矩阵的位置信息。
其中,在本公开的一个实施例之中,该预设阈值可以预先设置。
以及,在本公开的一个实施例之中,确定自信息量矩阵中元素值小于预设阈值的元素的方法可以包括:
构建一衰减函数δ(I i(p i)),使得该衰减系数与自信息量成反比,根据该衰减函数来确定自信息量矩 阵置0元素。
其中,在本公开的一个实施例之中,衰减系数可以服从玻尔兹曼分布,该衰减函数例如可以为:
Figure PCTCN2021098711-appb-000009
T可以看做一软门限,当T值较小时,自信息量矩阵中元素值很小的元素被置0,当T值趋近于无穷时,自信息量矩阵中元素值都被等概率置0,即随机置0。
在本公开的一个实施例之中,该衰减函数中的T取值应当相对较小。
以及,在本公开的一个实施例之中,基于衰减系数与自信息量成反比,则当自信息量大于预设阈值时,对应衰减率较小,则保留,当自信息量不大于预设阈值时,对应衰减率较大,则将其置0,以此得到稀疏自信息量矩阵。以及,应当认识到,在本公开的一个实施例之中,该稀疏自信息量矩阵的大小与CSI图像信息矩阵的大小相同。
由此可知,本公开实施例中,通过将自信息量矩阵中元素值小于预设阈值的元素用0代替,则可去除自信息量矩阵中的冗余信息,从而后续在基于自信息量矩阵对CSI图像矩阵进行压缩量化时,可以方便压缩,确保了压缩效率。
进一步地,在本公开的一个实施例之中,在确定出稀疏自信息量矩阵后,还可以确定出该稀疏自信息量矩阵中的非0元素在稀疏自信息量矩阵的位置信息,以便后续UE可以将该位置信息发送至基站,使得基站能够根据该位置信息重建CSI信息矩阵,确保了重建出的CSI信息矩阵的准确度。
步骤205、利用第三卷积层对稀疏自信息量矩阵进行特征处理,得到n个第二特征图。
其中,在本公开的一个实施例之中,该第三卷积层可以为无梯度更新的卷积层。
以及,在本公开的一个实施例之中,该n个第二特征图可以为自信息量矩阵对应的n维特征图。示例的,在本公开的一个实施例之中,n可以等于64,即:利用第三卷积层对自信息量矩阵进行特征处理,得到该自信息量矩阵对应的64维第二特征图。
需要说明的是,在本公开的一个实施例之中,第一卷积层和第三卷积层的结构原理可以相同,均用于对输入矩阵进行特征分析,以得到输入矩阵对应的n个特征图。
步骤206、利用第四卷积层基于n个第一特征图和n个第二特征图确定出稀疏CSI信息矩阵,以及稀疏CSI信息矩阵中的非0元素在所述稀疏CSI信息矩阵的位置信息。
其中,在本公开的一个实施例之中,该第四卷积层可以为带有梯度更新的卷积层。以及,示例的,在本公开的一个实施例之中,该第四卷积层的卷积核大小可以为c×n×s×s,卷积步长可以为L。示例的,在本公开的一个实施例之中,该第四卷积层的卷积核大小可以为2×64×3×3,卷积步长可以为1。
需要说明的是,在本公开的一个实施例之中,稀疏CSI信息矩阵与CSI信息矩阵的大小相同、维度相同。
则由上述内容可知,通过执行上述步骤201-206即可基于CSI信息矩阵的自信息量对CSI信息矩阵中的元素进行筛选,得到稀疏CSI信息矩阵。
在一些实施例中,上述的方案还可以包括:
步骤103、基于稀疏CSI信息矩阵中的元素确定反馈信息,并将反馈信息发送至基站。
图4为本公开一个实施例所提供的确定反馈信息的流程示意图,如图4所示,该方法可以包括:
步骤401、对稀疏CSI信息矩阵中的元素进行压缩量化,得到压缩码字。
其中,在本公开的一个实施例之中,对稀疏CSI信息矩阵中的元素进行压缩量化的方法可以包括:
步骤A、将稀疏CSI信息矩阵中各个元素值按照自信息量进行从大到小的排序。
步骤B、从排序后的元素值中选择出前M个元素值压缩量化构成压缩码字,其中,一个元素值对应一个码字。
其中,在本公开的一个实施例之中,确定M的方法可以包括:
步骤B1、确定压缩率σ。
其中,在本公开的一个实施例之中,该压缩率σ可以是基于用户需求确定的。示例的,在本公开的一个实施例之中,可以确定压缩率σ等于1/4。
步骤B2、基于公式三计算压缩码字的码数M,所述公式三包括:
Figure PCTCN2021098711-appb-000010
其中,a指示传输一个码字所需的比特数,k指示传输一个码字对应元素的位置信息所需的比特数,N c指示基站所设置的天线数,N t指示每根天线数对应的子载波数。
以及,需要说明的是,在本公开的一个实施例之中,由于UE向基站发送的反馈信息中并非是仅包括有压缩码字,还包括有位置信息,因此,公式三与现有技术中压缩率σ的公式有所不同,公式三中还包括有指示传输一个码字的位置信息所需的比特数k。
进一步地,在本公开的一个实施例中,步骤B中可以采用的经典Lloyd算法来压缩量化前M个元素值,以及,可以将该M个元素值进行8比特量化。
步骤C、将前M个元素值中各个元素值在稀疏CSI信息矩阵中的位置信息确定为被压缩的元素值在稀疏CSI信息矩阵中的位置信息。
步骤402、确定被压缩的元素在稀疏CSI信息矩阵中的位置信息,将压缩码字和位置信息确定为反馈信息。
综上所述,在本公开实施例提供的信息反馈方法之中,UE会对CSI信息矩阵中的元素进行筛选得到稀疏CSI信息矩阵,具体的,在确定稀疏CSI信息矩阵时,会先对CSI信息矩阵对应的CSI图像信息矩阵进行图像划分得到多个子图像信息矩阵,之后,会根据每个子图像信息矩阵和邻近子图像信息矩阵确定出每个子图像信息矩阵对应的分布函数,再基于分布函数确定出每个子图像信息矩阵对应的自信息量,最后,基于自信息量对冗余信息进行筛选以确定出稀疏CSI信息矩阵。以及,进一步地,UE会根据确定出的稀疏CSI信息矩阵中的元素确定反馈信息,并将反馈信息发送至基站,以使得基站基于该反馈信息重建CSI信息矩阵。
由此可知,本公开实施例中,在确定分布函数时,会考虑到每个子图像矩阵与邻近子图像矩阵之间的结构相关性,则当后续基于分布函数确定稀疏CSI信息矩阵时,不会破坏原始CSI信息矩阵的结构性,从而后续基站利用根据稀疏CSI信息矩阵确定出的反馈信息重建CSI信息矩阵时,可以确保重建的CSI信息矩阵的准确度。
同时,由于本公开实施例中的稀疏CSI信息矩阵是删除了冗余信息之后的矩阵,则可以确保后续对于稀疏CSI信息矩阵操作的便捷性,且降低了开销。
此外,本公开实施例中UE所确定出的反馈信息包括有被压缩的元素在稀疏CSI信息矩阵中的位置信息,从而后续基站可以根据被压缩的元素在稀疏CSI信息矩阵中的位置信息准确重建CSI信息矩阵,则进一步确保了CSI信息矩阵的重建准确度。
图5为本公开实施例所提供的一种信息反馈方法的流程示意图,应用于基站,如图5所示,该信息反馈方法可以包括以下步骤:
步骤501、获取UE发送的反馈信息,并基于反馈信息确定预备CSI信息矩阵。
其中,在本公开的一个实施例之中,该反馈信息可以包括:对CSI信息矩阵对应的稀疏CSI信息矩阵中的元素进行压缩量化后得到的压缩码字,以及被压缩的元素在所述稀疏CSI信息矩阵中的位置信息。
基于此,在本公开的一个实施例之中,基于反馈信息确定预备CSI信息矩阵可以包括:
步骤(1)、对压缩码字进行反量化得到反量化后的码字。
其中,在本公开的一个实施例之中,通过对压缩码字进行反量化则可确定出每一码字对应的元素值。
步骤(2)、构建一第二空矩阵,第二空矩阵的大小与CSI信息矩阵的大小相同。
步骤(3)、基于位置信息将反量化后的码字填充至第二空矩阵中。
其中,在本公开的一个实施例之中,基于位置信息将反量化后的码字填充至第二空矩阵中,可以包括:
若某一反量化后的码字对应的位置信息为:第W行、第Q列,则可将该反量化后的码字填充至第二空矩阵的第W行、第Q列处。
步骤(4)、计算反量化后的码字的均值,将均值填充至所述第二空矩阵的其他位置,得到预备CSI 信息矩阵。
其中,由于上述步骤401中仅是对稀疏CSI信息矩阵中的M个码字进行了压缩量化,因此,反量化后的码字也应当仅为M个元素。由此,在本公开的一个实施例之中,可能会出现反量化后的码字数量小于第二空矩阵的位置数的情况。也即是,反量化后的码字无法将第二空矩阵填满。
基于此,在本公开的一个实施例之中,可以将反量化后的码字的均值填充在第二空矩阵的其他位置处,以填满第二空矩阵得到预备CSI信息矩阵。
步骤502、基于预备CSI信息矩阵重建CSI信息矩阵。
其中,在本公开的一个实施例之中,基于预备CSI信息矩阵确定CSI信息矩阵,可以包括:将预备CSI信息矩阵输入至预先训练好的卷积结构中,以输出重建CSI信息矩阵。其中,需要说明的是,本公开的一个实施例之中,该卷积结构可以包括依次连接的第五卷积层、第六卷积层、第七卷积层、第八卷积层、第九卷积层、以及第十卷积层;
以及,本公开的一个实施例之中,第五卷积层的输出端还与八卷积层的输入端连接,第八卷积层的输出端还与第十卷积层的输入端连接。
进一步地,在本公开的一个实施例之中,第五层的卷积核可以为R 1×c×s×s,第六层卷积核可以为R 2×R 1×s×s,第七层卷积核可以为c×R 2×s×s,第八层卷积核为R 1×c×s×s,第九层卷积核为R 2×R 1×s×s,第十层卷积核为c×R 2×s×s,以及,第五卷积层、第六卷积层、第七卷积层、第八卷积层、第九卷积层、以及第十卷积层的步长可以相同。
示例的,在本公开的一个实施例之中,第四层的卷积核可以为8×2×3×3,第五层卷积核可以为16×8×3×3,第六层卷积核可以为2×16×3×3,第七层卷积核为8×2×3×3,第八层卷积核为16×8×3×3,第九层卷积核为2×16×3×3,以及,第五卷积层、第六卷积层、第七卷积层、第八卷积层、第九卷积层、以及第十卷积层的步长可以均为1。
以及,本公开的一个实施例之中,卷积结构的卷积操作可以定义为:
Figure PCTCN2021098711-appb-000011
其中y d,i,j为卷积结构输出的第(d,i,j)个元素,d为输出通道数,i为y d,i,j在重建CSI信息矩阵中的行数,j为y d,i,j在重建CSI信息矩阵中的列数,W d,c,h,w为卷积核权重矩阵W中第(d,c,h,w)个元素,c为输入通道数,h为卷积核的长度,w为卷积核的宽度,b d为卷积核偏置b中第d个元素,
Figure PCTCN2021098711-appb-000012
为卷积输入的第(c,i×s 1+h,j×s 2+w)个元素,i为
Figure PCTCN2021098711-appb-000013
在预备CSI信息矩阵中的行数,j为
Figure PCTCN2021098711-appb-000014
在预备CSI信息矩阵中的列数,s 1和s 2为卷积步长,记为(s 1,s 2),其中,s 1为卷积核横向移动步长,s 2为卷积核纵向移动步长。其中,卷积层每层的激活函数选取的为Leakyrelu函数,定义为:
Figure PCTCN2021098711-appb-000015
进一步地,在本公开的一个实施例之中,重建CSI信息矩阵具体包括对应于CSI信息矩阵实部的重建CSI信息矩阵,和对应于CSI信息矩阵虚部的重建CSI信息矩阵。基于此,在确定出对应于CSI信息矩阵实部的重建CSI信息矩阵,和对应于CSI信息矩阵虚部的重建CSI信息矩阵后,还可以将对应于CSI信息矩阵实部的重建CSI信息矩阵和对应于CSI信息矩阵虚部的重建CSI信息矩阵组合成复数矩阵,再对该复数矩阵进行二维离散傅里叶反变换获得空频域CSI信息矩阵。
综上所述,在本公开实施例提供的信息反馈方法之中,UE会对CSI信息矩阵中的元素进行筛选得到稀疏CSI信息矩阵,具体的,在确定稀疏CSI信息矩阵时,会先对CSI信息矩阵对应的CSI图像信息矩阵进行图像划分得到多个子图像信息矩阵,之后,会根据每个子图像信息矩阵和邻近子图像信息矩阵确定出每个子图像信息矩阵对应的分布函数,再基于分布函数确定出每个子图像信息矩阵对应的自信息量,最后,基于自信息量对冗余信息进行筛选以确定出稀疏CSI信息矩阵。以及,进一步地,UE会根据确定出的稀疏CSI信息矩阵中的元素确定反馈信息,并将反馈信息发送至基站,以使得基站基于该反馈信息重建CSI信息矩阵。
由此可知,本公开实施例中,在确定分布函数时,会考虑到每个子图像矩阵与邻近子图像矩阵之间 的结构相关性,则当后续基于分布函数确定稀疏CSI信息矩阵时,不会破坏原始CSI信息矩阵的结构性,从而后续基站利用根据稀疏CSI信息矩阵确定出的反馈信息重建CSI信息矩阵时,可以确保重建的CSI信息矩阵的准确度。
同时,由于本公开实施例中的稀疏CSI信息矩阵是删除了冗余信息之后的矩阵,则可以确保后续对于稀疏CSI信息矩阵操作的便捷性,且降低了开销。
此外,本公开实施例中UE所确定出的反馈信息包括有被压缩的元素在稀疏CSI信息矩阵中的位置信息,从而后续基站可以根据被压缩的元素在稀疏CSI信息矩阵中的位置信息准确重建CSI信息矩阵,则进一步确保了CSI信息矩阵的重建准确度。
图6为本公开实施例提供的一种信息反馈装置的结构示意图,配置于UE,如图6所示,可以包括:
第一获取模块,用于获取信道状态信息CSI信息矩阵。
筛选模块,用于基于所述CSI信息矩阵的自信息量对所述CSI信息矩阵中的元素进行筛选,得到稀疏CSI信息矩阵;
特征编码器,用于基于所述稀疏CSI信息矩阵中的元素确定反馈信息,并将所述反馈信息发送至基站。
综上所述,在本公开实施例提供的信息反馈装置之中,UE会对CSI信息矩阵中的元素进行筛选得到稀疏CSI信息矩阵,具体的,在确定稀疏CSI信息矩阵时,会先对CSI信息矩阵对应的CSI图像信息矩阵进行图像划分得到多个子图像信息矩阵,之后,会根据每个子图像信息矩阵和邻近子图像信息矩阵确定出每个子图像信息矩阵对应的分布函数,再基于分布函数确定出每个子图像信息矩阵对应的自信息量,最后,基于自信息量对冗余信息进行筛选以确定出稀疏CSI信息矩阵。以及,进一步地,UE会根据确定出的稀疏CSI信息矩阵中的元素确定反馈信息,并将反馈信息发送至基站,以使得基站基于该反馈信息重建CSI信息矩阵。
由此可知,本公开实施例中,在确定分布函数时,会考虑到每个子图像矩阵与邻近子图像矩阵之间的结构相关性,则当后续基于分布函数确定稀疏CSI信息矩阵时,不会破坏原始CSI信息矩阵的结构性,从而后续基站利用根据稀疏CSI信息矩阵确定出的反馈信息重建CSI信息矩阵时,可以确保重建的CSI信息矩阵的准确度。
同时,由于本公开实施例中的稀疏CSI信息矩阵是删除了冗余信息之后的矩阵,则可以确保后续对于稀疏CSI信息矩阵操作的便捷性,且降低了开销。
此外,本公开实施例中UE所确定出的反馈信息包括有被压缩的元素在稀疏CSI信息矩阵中的位置信息,从而后续基站可以根据被压缩的元素在稀疏CSI信息矩阵中的位置信息准确重建CSI信息矩阵,则进一步确保了CSI信息矩阵的重建准确度。
可选的,在本公开的一个实施例之中,所述第一获取模块还用于:将获取到的所述CSI信息矩阵从空频域变换到角度时延域,其中,所述CSI信息矩阵包括实部和虚部。
可选的,在本公开的另一个实施例之中,所述筛选模块,还用于:
将所述CSI信息矩阵映射为CSI图像信息矩阵;
利用第一卷积层对所述CSI图像信息矩阵进行特征处理,得到n个第一特征图,n为整数,n≥1;
利用第二卷积层确定所述CSI图像信息矩阵对应的自信息量矩阵,所述自信息量矩阵的大小与所述CSI图像信息矩阵的大小相同;
将所述自信息量矩阵中元素值小于预设阈值的元素用0代替得到稀疏自信息量矩阵,并确定所述稀疏自信息量矩阵中的非0元素在所述稀疏自信息量矩阵的位置信息;
利用第三卷积层对所述稀疏自信息量矩阵进行特征处理,得到n个第二特征图;
利用第四卷积层基于所述n个第一特征图和所述n个第二特征图确定出稀疏CSI信息矩阵,以及所述稀疏CSI信息矩阵中的非0元素在所述稀疏CSI信息矩阵的位置信息。
可选的,在本公开的另一个实施例之中,所述筛选模块,还用于:
在所述CSI图像信息矩阵外周添加0,以扩充所述CSI图像信息矩阵得到扩充CSI图像信息矩阵,将所述扩充CSI图像信息矩阵划分为m个子图像矩阵;其中,m为整数,m≥2,且m与所述CSI图像 信息矩阵中所包括的元素数量一致;
利用第二卷积层基于每个子图像矩阵以及每个子图像矩阵的邻近子图像矩阵计算出每个子图像矩阵的分布函数,基于所述分布函数计算出每个子图像矩阵的自信息量,并基于每个子图像矩阵的自信息量组成所述自信息量矩阵。
可选的,在本公开的另一个实施例之中,所述筛选模块,还用于:
选取输入至第二卷积层的第i个子图像矩阵p i,i为整数,i≥1;
确定所述p i的邻近子图像矩阵p i’,得到所述p i对应的邻近子图像集合Z i;其中,所述邻近子图像集合Z i包括:以所述p i为中心、以曼哈顿半径R为半径的圆内的(2R+1) 2个邻近子图像矩阵p i’;
通过公式一确定所述p i的分布函数f i(p i),所述公式一包括:
Figure PCTCN2021098711-appb-000016
其中,
Figure PCTCN2021098711-appb-000017
h表示p i与p i’之间的带宽。
可选的,在本公开的另一个实施例之中,所述筛选模块,还用于:
利用公式二基于第i个子图像矩阵的分布函数f i(p i)计算出所述第i个子图像矩阵的自信息量I i(p i),所述公式二包括:
I i(p i)=-log(f i(p i))。
可选的,在本公开的另一个实施例之中,所述筛选模块,还用于:
建立一第一空矩阵,所述第一空矩阵的大小与所述CSI图像信息矩阵的大小相同;
按照每个子图像矩阵在所述扩充CSI图像信息矩阵的位置将每个子图像矩阵对应的自信息量填充至所述第一空矩阵中,以组成自信息量矩阵。
可选的,在本公开的另一个实施例之中,所述第一卷积层和所述第四卷积层均为带有梯度更新的卷积层。
可选的,在本公开的另一个实施例之中,所述装置还用于:对所述第一卷积层和所述第四卷积层进行训练。
可选的,在本公开的另一个实施例之中,所述特征编码器,还包括:
压缩量化编码器,用于对所述稀疏CSI信息矩阵中的元素进行压缩量化,得到压缩码字;
确定单元,还用于确定被压缩的元素在所述稀疏CSI信息矩阵中的位置信息,将所述压缩码字和所述位置信息确定为所述反馈信息。
可选的,在本公开的一个实施例之中,所述装置还包括连接于筛选模块和特征编码器之间的第一维度映射模块,用于当筛选模块输出的稀疏CSI信息矩阵的维度不适用于特征编码器的维度时,将筛选模块输出的稀疏CSI信息矩阵的维度映射为适用于特征编码器的维度。
图7为本公开实施例提供的一种信息反馈装置的结构示意图,配置于基站,如图7所示,可以包括:
第二获取模块,用于获取UE发送的反馈信息,并基于所述反馈信息确定预备CSI信息矩阵;
重建模块,用于基于所述预备CSI信息矩阵重建CSI信息矩阵。
综上所述,在本公开实施例提供的信息反馈装置之中,UE会对CSI信息矩阵中的元素进行筛选得到稀疏CSI信息矩阵,具体的,在确定稀疏CSI信息矩阵时,会先对CSI信息矩阵对应的CSI图像信息矩阵进行图像划分得到多个子图像信息矩阵,之后,会根据每个子图像信息矩阵和邻近子图像信息矩阵确定出每个子图像信息矩阵对应的分布函数,再基于分布函数确定出每个子图像信息矩阵对应的自信息量,最后,基于自信息量对冗余信息进行筛选以确定出稀疏CSI信息矩阵。以及,进一步地,UE会根据确定出的稀疏CSI信息矩阵中的元素确定反馈信息,并将反馈信息发送至基站,以使得基站基于该反馈信息重建CSI信息矩阵。
由此可知,本公开实施例中,在确定分布函数时,会考虑到每个子图像矩阵与邻近子图像矩阵之间 的结构相关性,则当后续基于分布函数确定稀疏CSI信息矩阵时,不会破坏原始CSI信息矩阵的结构性,从而后续基站利用根据稀疏CSI信息矩阵确定出的反馈信息重建CSI信息矩阵时,可以确保重建的CSI信息矩阵的准确度。
同时,由于本公开实施例中的稀疏CSI信息矩阵是删除了冗余信息之后的矩阵,则可以确保后续对于稀疏CSI信息矩阵操作的便捷性,且降低了开销。
此外,本公开实施例中UE所确定出的反馈信息包括有被压缩的元素在稀疏CSI信息矩阵中的位置信息,从而后续基站可以根据被压缩的元素在稀疏CSI信息矩阵中的位置信息准确重建CSI信息矩阵,则进一步确保了CSI信息矩阵的重建准确度。
可选的,在本公开的一个实施例之中,所述反馈信息包括:对所述CSI信息矩阵对应的稀疏CSI信息矩阵中的元素进行压缩量化后得到的压缩码字,以及被压缩的元素在所述稀疏CSI信息矩阵中的位置信息;
所述第二获取模块,还包括:
反量化器,用于对所述压缩码字进行反量化得到反量化后的码字;
空矩阵构建单元,用于构建一第二空矩阵,所述第二空矩阵的大小与所述CSI信息矩阵的大小相同;
插值单元,用于基于所述位置信息将所述反量化后的码字填充至所述第二空矩阵中;
均值填充单元,用于计算所述反量化后的码字的均值,将所述均值填充至所述第二空矩阵的其他位置,得到预备CSI信息矩阵。
可选的,在本公开的一个实施例之中,所述重建模块,还用于:
利用预先训练好的卷积结构获取所述预备CSI信息矩阵,以输出所述CSI信息矩阵。
可选的,在本公开的一个实施例之中,所述卷积结构包括依次连接的第五卷积层、第六卷积层、第七卷积层、第八卷积层、第九卷积层、以及第十卷积层;
其中,所述第五卷积层的输出端还与所述八卷积层的输入端连接,所述第八卷积层的输出端还与所述第十卷积层的输入端连接。
可选的,在本公开的一个实施例之中,所述装置还用于:对所述卷积结构进行训练。
可选的,在本公开的一个实施例之中,所述重建模块还包括与第十卷积层输出端连接的第二维度映射模块,用于将第十卷积层输出的矩阵的维度映射为原始CSI信息矩阵的维度。
可选的,在本公开的一个实施例之中,所述重建CSI信息矩阵包括对应于CSI信息矩阵实部的重建CSI信息矩阵和对应于CSI信息矩阵虚部的重建CSI信息矩阵;
所述重建模块,还用于将所述对应于CSI信息矩阵实部的重建CSI信息矩阵和所述对应于CSI信息矩阵虚部的重建CSI信息矩阵组合成复数矩阵,再对该复数矩阵进行二维离散傅里叶反变换获得空频域CSI信息矩阵。
此外,图8为本公开实施例提供的一种信息反馈系统模型的结构示意图,如图8所示,包括依次连接的第一获取模块、筛选模块、第一维度映射模块、特征编码器、第二获取模块、以及重建模块,其中,第一获取模块、筛选模块、第二维度映射模块、特征编码器、第二获取模块、以及重建模块的详细介绍可以参考上述描述,本公开实施例在此不做赘述。
综上所述,在本公开实施例提供的信息反馈系统模型之中,UE会对CSI信息矩阵中的元素进行筛选得到稀疏CSI信息矩阵,具体的,在确定稀疏CSI信息矩阵时,会先对CSI信息矩阵对应的CSI图像信息矩阵进行图像划分得到多个子图像信息矩阵,之后,会根据每个子图像信息矩阵和邻近子图像信息矩阵确定出每个子图像信息矩阵对应的分布函数,再基于分布函数确定出每个子图像信息矩阵对应的自信息量,最后,基于自信息量对冗余信息进行筛选以确定出稀疏CSI信息矩阵。以及,进一步地,UE会根据确定出的稀疏CSI信息矩阵中的元素确定反馈信息,并将反馈信息发送至基站,以使得基站基于该反馈信息重建CSI信息矩阵。
由此可知,本公开实施例中,在确定分布函数时,会考虑到每个子图像矩阵与邻近子图像矩阵之间 的结构相关性,则当后续基于分布函数确定稀疏CSI信息矩阵时,不会破坏原始CSI信息矩阵的结构性,从而后续基站利用根据稀疏CSI信息矩阵确定出的反馈信息重建CSI信息矩阵时,可以确保重建的CSI信息矩阵的准确度。
同时,由于本公开实施例中的稀疏CSI信息矩阵是删除了冗余信息之后的矩阵,则可以确保后续对于稀疏CSI信息矩阵操作的便捷性,且降低了开销。
此外,本公开实施例中UE所确定出的反馈信息包括有被压缩的元素在稀疏CSI信息矩阵中的位置信息,从而后续基站可以根据被压缩的元素在稀疏CSI信息矩阵中的位置信息准确重建CSI信息矩阵,则进一步确保了CSI信息矩阵的重建准确度。
进一步地,在本公开的一个实施例之中,可以对图8所示的信息反馈系统模型进行训练,其中,训练方法可以包括:
步骤一、先获取CSI信息矩阵样本集合,其中,该CSI信息矩阵样本集合可以包括训练样本、验证样本、以及测试样本。
示例的,在本公开的一个实施例之中,可以使用COST2100[7]信道模型,在5.3GHz室内微蜂窝场景产生150000个空频域CSI信息矩阵样本。其中,训练样本可以包括100000个,验证样本可以包括30000,测试样本可以包括20000。以及,在本公开的一个实施例中,训练信息反馈系统模型时的epoch=1000,优化器可以采用Adam优化器,学习率learning rate=0.001,训练信息反馈系统模型时的batch=200。
步骤二、采用上述图1至图5所示的方法基于CSI信息矩阵样本集合对该信息反馈系统模型进行训练,计算损失函数L。
其中,在本公开的一个实施例之中,损失函数L可以定义为:
Figure PCTCN2021098711-appb-000018
其中,N为训练样本数,
Figure PCTCN2021098711-appb-000019
为配置于基站的重建模块输出的重建CSI信息矩阵,H a为配置于UE的第一获取模块获取到的原始CSI信息矩阵。‖·‖为欧几里得范数。
步骤三、根据损失函数对信息反馈系统模型的参数进行更新。
其中,在本公开的一个实施例之中,信息反馈系统模型的参数可以包括各个卷积层的权重和偏置。
重复执行上述步骤一至步骤三,直至损失函数收敛,确定训练完毕。
综上所述,由于本公开实施例中是采用图1至图5所示的方法对信息反馈系统模型进行训练的,其中,基于图1至图5所示的方法中矩阵的冗余信息较少、开销较低,从而使得对于信息反馈系统模型的训练方法的复杂度也较低,收敛较快,则提高了训练方法的精度和效率。
此外,如图8所示,重建模块中的第五卷积层的输出端还与八卷积层的输入端连接,第八卷积层的输出端还与第十卷积层的输入端连接,则可防止重建模块训练时出现梯度消失的现象,确保了训练精度。
本公开实施例提供的计算机存储介质,存储有可执行程序;所述可执行程序被处理器执行后,能够实现如图1至图4或图5任一所示的方法。
为了实现上述实施例,本公开还提出一种计算机程序产品,包括计算机程序,所述计算机程序在被处理器执行时实现如图1至图4或图5任一所示的方法。
此外,为了实现上述实施例,本公开还提出一种计算机程序,该程序被处理器执行时,以实现如图1至图4或图5任一所示的方法。
图9是本公开一个实施例所提供的一种用户设备UE900的框图。例如,UE900可以是移动电话,计算机,数字广播终端设备,消息收发设备,游戏控制台,平板设备,医疗设备,健身设备,个人数字助理等。
参照图9,UE900可以包括以下至少一个组件:处理组件902,存储器904,电源组件906,多媒体组件908,音频组件910,输入/输出(I/O)的接口912,传感器组件913,以及通信组件916。
处理组件902通常控制UE900的整体操作,诸如与显示,电话呼叫,数据通信,相机操作和记录 操作相关联的操作。处理组件902可以包括至少一个处理器920来执行指令,以完成上述的方法的全部或部分步骤。此外,处理组件902可以包括至少一个模块,便于处理组件902和其他组件之间的交互。例如,处理组件902可以包括多媒体模块,以方便多媒体组件908和处理组件902之间的交互。
存储器904被配置为存储各种类型的数据以支持在UE900的操作。这些数据的示例包括用于在UE900上操作的任何应用程序或方法的指令,联系人数据,电话簿数据,消息,图片,视频等。存储器904可以由任何类型的易失性或非易失性存储设备或者它们的组合实现,如静态随机存取存储器(SRAM),电可擦除可编程只读存储器(EEPROM),可擦除可编程只读存储器(EPROM),可编程只读存储器(PROM),只读存储器(ROM),磁存储器,快闪存储器,磁盘或光盘。
电源组件906为UE900的各种组件提供电力。电源组件906可以包括电源管理系统,至少一个电源,及其他与为UE900生成、管理和分配电力相关联的组件。
多媒体组件908包括在所述UE900和用户之间的提供一个输出接口的屏幕。在一些实施例中,屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板,屏幕可以被实现为触摸屏,以接收来自用户的输入信号。触摸面板包括至少一个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界,而且还检测与所述触摸或滑动操作相关的唤醒时间和压力。在一些实施例中,多媒体组件908包括一个前置摄像头和/或后置摄像头。当UE900处于操作模式,如拍摄模式或视频模式时,前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件910被配置为输出和/或输入音频信号。例如,音频组件910包括一个麦克风(MIC),当UE900处于操作模式,如呼叫模式、记录模式和语音识别模式时,麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器904或经由通信组件916发送。在一些实施例中,音频组件910还包括一个扬声器,用于输出音频信号。
I/O接口912为处理组件902和外围接口模块之间提供接口,上述外围接口模块可以是键盘,点击轮,按钮等。这些按钮可包括但不限于:主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件913包括至少一个传感器,用于为UE900提供各个方面的状态评估。例如,传感器组件913可以检测到设备900的打开/关闭状态,组件的相对定位,例如所述组件为UE900的显示器和小键盘,传感器组件913还可以检测UE900或UE900一个组件的位置改变,用户与UE900接触的存在或不存在,UE900方位或加速/减速和UE900的温度变化。传感器组件913可以包括接近传感器,被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件913还可以包括光传感器,如CMOS或CCD图像传感器,用于在成像应用中使用。在一些实施例中,该传感器组件913还可以包括加速度传感器,陀螺仪传感器,磁传感器,压力传感器或温度传感器。
通信组件916被配置为便于UE900和其他设备之间有线或无线方式的通信。UE900可以接入基于通信标准的无线网络,如WiFi,2G或3G,或它们的组合。在一个示例性实施例中,通信组件916经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中,所述通信组件916还包括近场通信(NFC)模块,以促进短程通信。例如,在NFC模块可基于射频识别(RFID)技术,红外数据协会(IrDA)技术,超宽带(UWB)技术,蓝牙(BT)技术和其他技术来实现。
在示例性实施例中,UE900可以被至少一个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现,用于执行上述方法。
图10是本公开实施例所提供的一种基站1000的框图。例如,基站1000可以被提供为一基站。参照图10,基站1000包括处理组件1026,其进一步包括至少一个处理器,以及由存储器1032所代表的存储器资源,用于存储可由处理组件1022的执行的指令,例如应用程序。存储器1032中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外,处理组件1026被配置为执行指令,以执行上述方法前述应用在所述基站的任意方法,例如,如图1所示方法。
基站1000还可以包括一个电源组件1026被配置为执行基站1000的电源管理,一个有线或无线网络接口1050被配置为将基站1000连接到网络,和一个输入输出(I/O)接口1058。基站1000可以操作基 于存储在存储器1032的操作系统,例如Windows Server TM,Mac OS XTM,Unix TM,Linux TM,Free BSDTM或类似。
本领域技术人员在考虑说明书及实践这里公开的发明后,将容易想到本发明的其它实施方案。本公开旨在涵盖本发明的任何变型、用途或者适应性变化,这些变型、用途或者适应性变化遵循本发明的一般性原理并包括本公开未公开的本技术领域中的公知常识或惯用技术手段。说明书和实施例仅被视为示例性的,本公开的真正范围和精神由下面的权利要求指出。
应当理解的是,本公开并不局限于上面已经描述并在附图中示出的精确结构,并且可以在不脱离其范围进行各种修改和改变。本公开的范围仅由所附的权利要求来限制。

Claims (24)

  1. 一种信息反馈方法,其特征在于,应用于用户设备UE,包括:
    获取信道状态信息CSI信息矩阵;
    基于所述CSI信息矩阵的自信息量对所述CSI信息矩阵中的元素进行筛选,得到稀疏CSI信息矩阵;
    基于所述稀疏CSI信息矩阵中的元素确定反馈信息,并将所述反馈信息发送至基站。
  2. 如权利要求1所述的方法,其特征在于,所述获取CSI信息矩阵包括:
    将获取到的所述CSI信息矩阵从空频域变换到角度时延域,其中,所述CSI信息矩阵包括实部和虚部。
  3. 如权利要求1所述的方法,其特征在于,所述基于所述CSI信息矩阵的自信息量对所述CSI信息矩阵中的元素进行筛选,包括:
    将所述CSI信息矩阵映射为CSI图像信息矩阵;
    利用第一卷积层对所述CSI图像信息矩阵进行特征处理,得到n个第一特征图,n为整数,n≥1;
    利用第二卷积层确定所述CSI图像信息矩阵对应的自信息量矩阵,所述自信息量矩阵的大小与所述CSI图像信息矩阵的大小相同;
    将所述自信息量矩阵中元素值小于预设阈值的元素用0代替得到稀疏自信息量矩阵,并确定所述稀疏自信息量矩阵中的非0元素在所述稀疏自信息量矩阵的位置信息;
    利用第三卷积层对所述稀疏自信息量矩阵进行特征处理,得到n个第二特征图;
    利用第四卷积层基于所述n个第一特征图和所述n个第二特征图确定出稀疏CSI信息矩阵,以及所述稀疏CSI信息矩阵中的非0元素在所述稀疏CSI信息矩阵的位置信息。
  4. 如权利要求3所述的方法,其特征在于,所述利用第二卷积层确定所述CSI图像信息矩阵对应的自信息量矩阵,包括:
    在所述CSI图像信息矩阵外周添加0,以扩充所述CSI图像信息矩阵得到扩充CSI图像信息矩阵,将所述扩充CSI图像信息矩阵划分为m个子图像矩阵;其中,m为整数,m≥2,且m与所述CSI图像信息矩阵中所包括的元素数量一致;
    利用第二卷积层基于每个子图像矩阵以及每个子图像矩阵的邻近子图像矩阵计算出每个子图像矩阵的分布函数,基于所述分布函数计算出每个子图像矩阵的自信息量,并基于每个子图像矩阵的自信息量组成所述自信息量矩阵。
  5. 如权利要求4所述的方法,其特征在于,所述利用第二卷积层基于每个子图像矩阵以及每个子图像矩阵的邻近子图像矩阵计算出每个子图像矩阵的分布函数,包括:
    选取输入至第二卷积层的第i个子图像矩阵p i,i为整数,i≥1;
    确定所述p i的邻近子图像矩阵p i’,得到所述p i对应的邻近子图像集合Z i;其中,所述邻近子图像集合Z i包括:以所述p i为中心、以曼哈顿半径R为半径的圆内的(2R+1) 2个邻近子图像矩阵p i’;
    通过公式一确定所述p i的分布函数f i(p i),所述公式一包括:
    Figure PCTCN2021098711-appb-100001
    其中,
    Figure PCTCN2021098711-appb-100002
    h表示p i与p i’之间的带宽。
  6. 如权利要求4所述的方法,其特征在于,所述基于所述分布函数计算出每个子图像矩阵的自信息量,包括:
    利用公式二基于第i个子图像矩阵的分布函数f i(p i)计算出所述第i个子图像矩阵的自信息量I i(p i),所述公式二包括:
    I i(p i)=-log(f i(p i))。
  7. 如权利要求4所述的方法,其特征在于,所述基于每个子图像矩阵的自信息量组成所述自信息量矩阵,包括:
    建立一第一空矩阵,所述第一空矩阵的大小与所述CSI图像信息矩阵的大小相同;
    按照每个子图像矩阵在所述扩充CSI图像信息矩阵的位置将每个子图像矩阵对应的自信息量填充至所述第一空矩阵中,以组成自信息量矩阵。
  8. 如权利要求3所述的方法,其特征在于,所述第一卷积层和所述第四卷积层均为带有梯度更新的卷积层。
  9. 如权利要求8所述的方法,其特征在于,还包括:
    对所述第一卷积层和所述第四卷积层进行训练。
  10. 如权利要求1所述的方法,其特征在于,所述基于所述稀疏CSI信息矩阵中的元素确定反馈信息,包括:
    对所述稀疏CSI信息矩阵中的元素进行压缩量化,得到压缩码字;
    确定被压缩的元素在所述稀疏CSI信息矩阵中的位置信息,将所述压缩码字和所述位置信息确定为所述反馈信息。
  11. 如权利要求10所述的方法,其特征在于,所述对所述稀疏CSI信息矩阵中的元素值进行压缩量化,以得到压缩码字,包括:
    将所述稀疏CSI信息矩阵中各个元素值按照自信息量进行从大到小的排序;
    从排序后的元素值中选择出前M个元素值构成压缩码字,其中,一个元素值对应一个码字;以及
    将所述前M个元素值中各个元素值在所述稀疏CSI信息矩阵中的位置信息确定为所述被压缩的元素值在所述稀疏CSI信息矩阵中的位置信息。
  12. 如权利要求11所述的方法,其特征在于,确定所述M的方法包括:
    确定压缩率σ;
    基于公式三计算所述压缩码字的码数M,所述公式三包括:
    Figure PCTCN2021098711-appb-100003
    其中,a指示传输一个码字所需的比特数,k指示传输一个码字的位置信息所需的比特数,N c指示基站所设置的天线数,N t指示每根天线数对应的子载波数。
  13. 一种信息反馈方法,其特征在于,应用于基站,包括:
    获取UE发送的反馈信息,并基于所述反馈信息确定预备CSI信息矩阵;
    基于所述预备CSI信息矩阵重建CSI信息矩阵。
  14. 如权利要求13所述的方法,其特征在于,所述反馈信息包括:对所述CSI信息矩阵对应的稀疏CSI信息矩阵中的元素进行压缩量化后得到的压缩码字,以及被压缩的元素在所述稀疏CSI信息矩阵中的位置信息;
    所述基于所述反馈信息确定预备CSI信息矩阵,包括:
    对所述压缩码字进行反量化得到反量化后的码字;
    构建一第二空矩阵,所述第二空矩阵的大小与所述CSI信息矩阵的大小相同;
    基于所述位置信息将所述反量化后的码字填充至所述第二空矩阵中;
    计算所述反量化后的码字的均值,将所述均值填充至所述第二空矩阵的其他位置,得到预备CSI信息矩阵。
  15. 如权利要求13所述的方法,其特征在于,所述基于所述预备CSI信息矩阵确定CSI信息矩阵,包括:
    利用预先训练好的卷积结构获取所述预备CSI信息矩阵,以输出所述CSI信息矩阵。
  16. 如权利要求15所述的方法,其特征在于,所述卷积结构包括依次连接的第五卷积层、第六卷积层、第七卷积层、第八卷积层、第九卷积层、以及第十卷积层;
    其中,所述第五卷积层的输出端还与所述八卷积层的输入端连接,所述第八卷积层的输出端还与所述第十卷积层的输入端连接。
  17. 如权利要求15所述的方法,其特征在于,还包括:对所述卷积结构进行训练。
  18. 一种信息反馈装置,其特征在于,包括:
    第一获取模块,用于获取信道状态信息CSI信息矩阵;
    筛选模块,用于基于所述CSI信息矩阵的自信息量对所述CSI信息矩阵中的元素进行筛选,得到稀疏CSI信息矩阵;
    特征编码器,用于基于所述稀疏CSI信息矩阵中的元素确定反馈信息,并将所述反馈信息发送至基站。
  19. 一种信息反馈装置,其特征在于,包括:
    第二获取模块,用于获取UE发送的反馈信息,并基于所述反馈信息确定预备CSI信息矩阵;
    重建模块,用于基于所述预备CSI信息矩阵重建CSI信息矩阵。
  20. 一种用户设备,其特征在于,包括:收发器;存储器;处理器,分别与所述收发器及所述存储器连接,配置为通过执行所述存储器上的计算机可执行指令,控制所述收发器的无线信号收发,并能够实现权利要求1至12任一项所述的方法。
  21. 一种基站,其特征在于,包括:收发器;存储器;处理器,分别与所述收发器及所述存储器连接,配置为通过执行所述存储器上的计算机可执行指令,控制所述收发器的无线信号收发,并能够实现权利要求13至17任一项所述的方法。
  22. 一种信息反馈系统模型,其特征在于,所述系统模型至少包括如权利要求18和19所述的信息反馈装置。
  23. 一种用于对权利要求22所述的信息反馈系统模型的训练方法,其特征在于,包括:通过执行如权利要求1至12和/或13至17任一项所述的方法以对所述信息反馈系统模型进行训练。
  24. 一种计算机存储介质,其中,所述计算机存储介质存储有计算机可执行指令;所述计算机可执行指令被处理器执行后,能够实现权利要求1至12或13至17任一项所述的方法。
PCT/CN2021/098711 2021-06-07 2021-06-07 信息反馈方法、装置、用户设备、基站、系统模型及存储介质 WO2022256988A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
PCT/CN2021/098711 WO2022256988A1 (zh) 2021-06-07 2021-06-07 信息反馈方法、装置、用户设备、基站、系统模型及存储介质
EP21944499.9A EP4354755A1 (en) 2021-06-07 2021-06-07 Information feedback method and apparatus, user equipment, base station, system model and storage medium
CN202180001750.9A CN115917982A (zh) 2021-06-07 2021-06-07 信息反馈方法、装置、用户设备、基站、系统模型及存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/098711 WO2022256988A1 (zh) 2021-06-07 2021-06-07 信息反馈方法、装置、用户设备、基站、系统模型及存储介质

Publications (1)

Publication Number Publication Date
WO2022256988A1 true WO2022256988A1 (zh) 2022-12-15

Family

ID=84424642

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/098711 WO2022256988A1 (zh) 2021-06-07 2021-06-07 信息反馈方法、装置、用户设备、基站、系统模型及存储介质

Country Status (3)

Country Link
EP (1) EP4354755A1 (zh)
CN (1) CN115917982A (zh)
WO (1) WO2022256988A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150256244A1 (en) * 2014-03-07 2015-09-10 Samsung Electronics Co., Ltd. Apparatus and method for channel feedback in multiple input multiple output system
CN105981320A (zh) * 2014-02-19 2016-09-28 华为技术有限公司 一种信道状态信息反馈方法、装置、用户设备及基站
US20200067573A1 (en) * 2016-11-03 2020-02-27 Wisig Networks Private Limited Method and System for Enabling Explicit Channel State Information (CSI) Feedback in Multiple-Input-Multiple-Output (MIMO)
CN111464220A (zh) * 2020-03-10 2020-07-28 西安交通大学 一种基于深度学习的信道状态信息重建方法
US20210126692A1 (en) * 2018-01-22 2021-04-29 Lg Electronics Inc. Method for transmitting and receiving channel state information in wireless communication system and device therefor

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105981320A (zh) * 2014-02-19 2016-09-28 华为技术有限公司 一种信道状态信息反馈方法、装置、用户设备及基站
US20150256244A1 (en) * 2014-03-07 2015-09-10 Samsung Electronics Co., Ltd. Apparatus and method for channel feedback in multiple input multiple output system
US20200067573A1 (en) * 2016-11-03 2020-02-27 Wisig Networks Private Limited Method and System for Enabling Explicit Channel State Information (CSI) Feedback in Multiple-Input-Multiple-Output (MIMO)
US20210126692A1 (en) * 2018-01-22 2021-04-29 Lg Electronics Inc. Method for transmitting and receiving channel state information in wireless communication system and device therefor
CN111464220A (zh) * 2020-03-10 2020-07-28 西安交通大学 一种基于深度学习的信道状态信息重建方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
JI-YUAN PAN ; JIANG-SHE ZHANG: "Nonnegative matrix factorization with maximum self-information on basis components", INFORMATION TECHNOLOGY AND ARTIFICIAL INTELLIGENCE CONFERENCE (ITAIC), 2011 6TH IEEE JOINT INTERNATIONAL, IEEE, 20 August 2011 (2011-08-20), pages 128 - 132, XP031967821, ISBN: 978-1-4244-8622-9, DOI: 10.1109/ITAIC.2011.6030167 *

Also Published As

Publication number Publication date
EP4354755A1 (en) 2024-04-17
CN115917982A (zh) 2023-04-04

Similar Documents

Publication Publication Date Title
US20230019669A1 (en) Systems and methods for enhanced feedback for cascaded federated machine learning
CN104272791B (zh) 无线多媒体体验质量报告
US20220247469A1 (en) Method and device for transmitting channel state information
CN111435862B (zh) 信道状态信息csi报告的传输方法、终端及网络设备
US11456791B2 (en) Generic feedback to enable reciprocity and over the air calibration for advanced networks
WO2020238906A1 (zh) 信道状态信息csi报告的传输方法、终端及网络设备
US10826591B1 (en) Irregular grid sub sampling to reduce feedback overhead in a 5G or other next generation wireless network
WO2022121797A1 (zh) 一种传输数据的方法和装置
KR20230015269A (ko) 촬영룸 키오스크를 통해 촬영된 이미지들에 기초하여 사진 프레임을 생성하는 장치 및 방법
WO2023077297A1 (zh) 信息反馈方法及装置、存储介质
US10547359B2 (en) Calibration data
CN115280833A (zh) 信道状态信息的处理方法、装置及通信设备
WO2022256988A1 (zh) 信息反馈方法、装置、用户设备、基站、系统模型及存储介质
CN111615142B (zh) 信道状态信息csi报告的传输方法、终端及网络设备
CN110034802A (zh) 一种信息传输方法及装置
CN118056429A (zh) 虚拟信道样本的质量评估方法和设备
CN114157722A (zh) 一种数据传输方法及装置
WO2023006096A1 (zh) 一种通信方法及装置
WO2023078229A1 (zh) 信息上报方法、终端及网络侧设备
WO2023115254A1 (zh) 处理数据的方法及装置
WO2022261842A1 (zh) 预编码矩阵确定方法、装置、用户设备、基站及存储介质
WO2024065804A1 (zh) 数据压缩传输方法、装置、设备以及存储介质
CN111436080A (zh) 一种信息发送方法、终端设备及网络设备
WO2023240572A1 (zh) 一种信息传输方法、装置、通信设备及存储介质
WO2024046288A1 (zh) 通信方法和装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21944499

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18567355

Country of ref document: US

WWE Wipo information: entry into national phase

Ref document number: 2021944499

Country of ref document: EP

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2021944499

Country of ref document: EP

Effective date: 20240108