CN115189742A - Differential coding assisted machine learning channel state information feedback method and device - Google Patents

Differential coding assisted machine learning channel state information feedback method and device Download PDF

Info

Publication number
CN115189742A
CN115189742A CN202210805014.5A CN202210805014A CN115189742A CN 115189742 A CN115189742 A CN 115189742A CN 202210805014 A CN202210805014 A CN 202210805014A CN 115189742 A CN115189742 A CN 115189742A
Authority
CN
China
Prior art keywords
differential
csi
signal
downlink csi
enhanced
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.)
Pending
Application number
CN202210805014.5A
Other languages
Chinese (zh)
Inventor
卿朝进
刘文慧
叶青
王子龙
陈金良
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xihua University
Original Assignee
Xihua University
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 Xihua University filed Critical Xihua University
Priority to CN202210805014.5A priority Critical patent/CN115189742A/en
Publication of CN115189742A publication Critical patent/CN115189742A/en
Pending legal-status Critical Current

Links

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
    • H04B7/06—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
    • H04B7/0621—Feedback content
    • H04B7/0626—Channel coefficients, e.g. channel state information [CSI]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/0202—Channel estimation
    • H04L25/024—Channel estimation channel estimation algorithms
    • H04L25/0254—Channel estimation channel estimation algorithms using neural network algorithms
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Power Engineering (AREA)
  • Digital Transmission Methods That Use Modulated Carrier Waves (AREA)

Abstract

The invention discloses a differential coding assisted machine learning channel state information feedback method and a differential coding assisted machine learning channel state information feedback device, wherein the method comprises the following steps: a user side: the user terminal carries out real-valued, quantitative and digital modulation on the downlink CSI vector to obtain a modulated signal, and then carries out differential coding to obtain a differential coded signal and feeds the differential coded signal back to the base station; a base station end: a character-by-character differential decoding mode is adopted for the received signals to obtain differential decoding signals; and carrying out digital demodulation and dequantization processing on the differential decoding signal to obtain downlink CSI recovery information, and constructing an enhanced CSI network Enh-CsiNet to carry out enhancement processing on the downlink CSI recovery information to obtain the downlink CSI with enhanced final recovery precision. The invention avoids the situation that the user terminal sends guidance, thereby saving bandwidth resources, reducing the energy expenditure of the user terminal and simultaneously improving the reconstruction precision of the downlink CSI.

Description

Differential coding assisted machine learning channel state information feedback method and device
Technical Field
The present invention relates to the technical field of Channel State Information feedback of a large-scale mimo system in a frequency division duplex mode, and in particular, to a method and an apparatus for feeding back Channel State Information (CSI) of a machine learning assisted by differential coding.
Background
In order to reduce feedback overhead and thus reduce uplink bandwidth resource occupation and User Equipment (UE) end energy consumption, the existing methods mainly focus on compression feedback of downlink CSI. The CSI feedback method based on Compressed Sensing (CS) reduces feedback overhead by developing a signal sparse structure. However, the CS-based CSI feedback is based on a sparse downlink CSI assumption under a certain sparse basis, which sometimes does not conform to reality, resulting in a significant degradation of downlink CSI sparse reconstruction performance. Deep Learning (DL) is applied to CSI feedback, so that CSI feedback accuracy is improved while CSI feedback overhead is effectively reduced. The DL-based overlay CSI feedback method superimposes downlink CSI on an Uplink User Data sequence (UL-US) and then feeds back the Uplink User Data sequence to a Base Station (BS) end, so that extra Uplink bandwidth resources occupied by feeding back downlink CSI are avoided, and the spectrum efficiency is improved. However, there is superimposed interference in superimposing CSI feedback, which reduces CSI feedback accuracy. Furthermore, in order for the BS side to demodulate the UL-US, the UE side generally needs to send a pilot for uplink channel estimation to the BS. Inevitably occupying uplink bandwidth resources and consuming UE energy. Fortunately, the machine learning CSI feedback method assisted by differential coding avoids the guiding overhead for uplink channel estimation in a differential mode, thereby reducing the occupation of uplink bandwidth resources and saving the energy consumption of a UE (user equipment) end, and the introduction of a neural network further improves the reconstruction precision of downlink CSI.
Disclosure of Invention
The invention aims to provide a differential coding assisted machine learning channel state information feedback method and device, which can save uplink bandwidth resources and reduce energy consumption of a user terminal and improve reconstruction accuracy of downlink CSI (channel state information) compared with the CSI feedback of the existing uplink pilot channel estimation.
The differential coding assisted machine learning channel state information feedback method comprises the following steps:
user side processing:
s1, enabling downlink CSI vectors
Figure BDA0003736677030000011
Carrying out real-valued, quantized and digital modulation to obtain a modulated signal s, carrying out differential coding to obtain a differential coded signal z, carrying out OFDM modulation on the differential coded signal z, and feeding back to the base station.
The real-valued transformation is to transform the downlink CSI vector h into real-valued CSI information
Figure BDA0003736677030000012
The quantization is to real-value CSI information
Figure BDA0003736677030000013
A-bit quantization is carried out to obtain the bit stream information of real part and imaginary part
Figure BDA0003736677030000014
The digital modulation is to convert bit stream information
Figure BDA0003736677030000015
Performing digital modulation to form modulated signal
Figure BDA0003736677030000016
Wherein the content of the first and second substances,
Figure BDA0003736677030000021
is a modulation order;
the differential coding is to modulate a signal s = [ s ] 1 ,s 2 ,...,s N ] T According to the rule
Figure BDA0003736677030000022
Carrying out differential coding to obtain a differential coded signal z = [) 1 ,z 2 ,...z N ] T ；
Wherein N =2aM γ; UE side and BS side agree on z at the same time 0 =1, then z does not have to be transmitted 0 The differential coding signal z is a codeAbandon z 0 The latter differentially encoded signal.
Base station end processing:
s2, the base station side carries out Orthogonal Frequency Division Multiplexing (OFDM) demodulation to obtain a received signal
Figure BDA0003736677030000023
And a differential decoding signal is obtained by adopting a character-by-character differential decoding mode for the received signal r
Figure BDA0003736677030000024
The differential decoding is realized by solving
Figure BDA0003736677030000025
Obtaining;
wherein the content of the first and second substances,
Figure BDA0003736677030000026
represents the ith decision value; r is a radical of hydrogen i 、r i-1 Respectively represent the ith element and the i-1 th element in the r; s t Represented in a constellation set
Figure BDA0003736677030000027
The upper mark indicates the conjugate operation, re [ ·]Representing the real part, max [. Cndot]Indicating a max operation.
S3, decoding the differential signal
Figure BDA0003736677030000028
Performing digital demodulation and dequantization to obtain downlink CSI recovery information
Figure BDA0003736677030000029
And constructing an enhanced CSI network Enh-CsiNet to recover information of downlink CSI
Figure BDA00037366770300000210
Enhancing to obtain downlink with enhanced recovery precision
Figure BDA00037366770300000211
The digital demodulation is to differentially decode the signals
Figure BDA00037366770300000212
Performing demodulation processing to obtain bit stream information
Figure BDA00037366770300000213
The dequantization process is based on the obtained bit stream information
Figure BDA00037366770300000214
Corresponding a-bit dequantization operation is carried out to obtain the recovery information of the downlink CSI
Figure BDA00037366770300000215
The enhanced CSI network Enh-CsiNEt of step S3 further includes:
an input layer, a hidden layer containing a Leaky Relu activation function, and an output layer containing a linear activation function. The number of nodes of the input layer, the hidden layer and the output layer is respectively 2M, nM and 2M, and n represents hidden layer node coefficients determined according to engineering presetting;
constructing a training data set
Figure BDA00037366770300000216
Training the enhanced CSI network Enh-CsiNet to obtain network parameters of the enhanced CSI network Enh-CsiNet;
the training label
Figure BDA00037366770300000217
The downlink CSI vector h is obtained by real-valued processing, that is:
h label ＝[Re(h 1 ),Im(h 1 ),…,Re(h M ),Im(h M )] T
wherein Im (·) represents taking the imaginary part, [ ·] T Representing a transposition operation;
when in online operation, the recovery information of the downlink CSI is transmitted
Figure BDA00037366770300000218
Inputting an enhanced CSI network Enh-CsiNet to obtain downlink with enhanced recovery precision
Figure BDA0003736677030000031
The invention provides a differential coding assisted machine learning channel state information feedback device, a user end device comprises:
the modulation module is used for obtaining a modulated signal S according to the design of the S1;
the differential coding module is used for obtaining a differential coding signal z according to the design in the S1;
the differential coding module is connected behind the modulation signal acquisition module.
The base station side device includes:
a differential decoding module for obtaining differential decoding signal according to the design of S2
Figure BDA0003736677030000032
A precision enhancing module for obtaining the downlink with enhanced recovery precision according to the design of S3
Figure BDA0003736677030000033
The modulation module includes:
quantization unit for real value CSI information
Figure BDA0003736677030000034
A-bit quantization is carried out to obtain real and imaginary bit stream information
Figure BDA0003736677030000035
A modulation unit for converting the bit stream information
Figure BDA0003736677030000036
Performing digital modulation to form modulated signal
Figure BDA0003736677030000037
The precision enhancement module includes:
a demodulation unit for differentially decoding the signals
Figure BDA0003736677030000038
Performing demodulation processing to obtain bit stream information
Figure BDA0003736677030000039
A dequantizing unit for dequantizing the bit stream information based on the obtained bit stream information
Figure BDA00037366770300000310
Performing corresponding a-bit dequantization operation to obtain recovery information of downlink CSI
Figure BDA00037366770300000311
A network enhancing unit for recovering the downlink CSI
Figure BDA00037366770300000312
Inputting the enhanced CSI network Enh-CsiNet to obtain the downlink with enhanced recovery precision
Figure BDA00037366770300000313
The invention utilizes differential coding to assist in CSI feedback, further improves the reconstruction precision of the downlink CSI by the aid of the light-quantization enhanced CSI network Enh-CsiNet, feeds back the differential coding signal subjected to differential coding to the base station through the OFDM modulation technology by combining the advantages of the quantization coding, the digital modulation and the DL technology, obtains the downlink CSI recovery information by the aid of differential decoding and digital demodulation and de-quantization processing at the base station end, further improves the reconstruction precision of the downlink CSI by the aid of the light-quantization enhanced CSI network Enh-CsiNet, greatly saves uplink bandwidth resources and UE end energy while avoiding the UE end from sending guidance for the BS end to carry out channel estimation, and ensures the reconstruction precision of the CSI.
Compared with the CSI feedback of the existing uplink pilot channel estimation, the method of the invention utilizes the differential coding technology to save uplink bandwidth resources and reduce energy consumption of a user terminal, and utilizes a light neural network to improve the reconstruction precision of downlink CSI.
Drawings
FIG. 1 is a schematic overall flow diagram of the present invention;
fig. 2 is a flow chart of a ue according to the present invention;
fig. 3 is a flow chart of a base station according to the present invention.
Detailed Description
The present invention is described in detail with reference to the following embodiments and drawings, but it should be understood that the embodiments and drawings are only for illustrative purposes and are not intended to limit the scope of the present invention. All reasonable variations and combinations that fall within the spirit of the invention are intended to be within the scope of the invention.
Referring to fig. 1, the specific differential coding assisted machine learning channel state information feedback method includes:
a1 Referring to fig. 2, downlink CSI vector h is modulated into modulated signal s by quantization and digital modulation, and then differentially encoded to obtain differentially encoded signal z, and the differentially encoded signal z is fed back to the base station after OFDM modulation.
Among these, some more specific embodiments are:
the real-valued transformation is to transform the downlink CSI vector h into real-valued CSI information
Figure BDA0003736677030000041
Namely:
Figure BDA0003736677030000042
wherein Re (-) represents the real part, im (-) represents the imaginary part, [. Cndot. ]] T Representing a transposition operation;
the quantization refers to the information of the real-value CSI
Figure BDA0003736677030000043
A-bit quantization is carried out to obtain the real and imaginary part bit stream information of the downlink CSI vector h
Figure BDA0003736677030000044
The digital modulation is to convert bit stream information
Figure BDA0003736677030000045
Performing digital modulation to form modulated signal
Figure BDA0003736677030000046
Wherein
Figure BDA0003736677030000047
Is the modulation order (positive integer);
the differential coding is to the modulated signal s = [ s ] 1 ,s 2 ,...,s N ] T According to the rule
Figure BDA0003736677030000048
Carrying out differential coding to obtain a differential coded signal z = [ z ] 1 ,z 2 ,...z N ] T ；
Wherein N =2aM γ; the differential coding signal z is agreed by the UE end and the BS end at the same time 0 =1, abandon z 0 Differential coded signal after = 1.
a2 Referring to fig. 3, the base station performs OFDM demodulation to obtain a received signal
Figure BDA0003736677030000049
And a differential decoding signal is obtained by adopting a character-by-character differential decoding mode for the received signal r
Figure BDA00037366770300000410
The differential decoding signal is obtained by solving
Figure BDA00037366770300000411
Obtaining;
wherein the content of the first and second substances,
Figure BDA00037366770300000412
represents the ith decision value; r is i 、r i-1 Respectively represent the i-th element and the i-1-th element in r; s t Represented in a constellation set
Figure BDA00037366770300000415
The upper mark indicates the conjugate operation, re [ ·]Representing the real part, max [. Cndot]Indicating a max operation.
a3 Will differentially decode the signal
Figure BDA00037366770300000413
Performing digital demodulation and dequantization to obtain downlink CSI recovery information
Figure BDA00037366770300000414
And constructing an enhanced CSI network Enh-CsiNet to recover information of downlink CSI
Figure BDA0003736677030000051
Enhancing to obtain downlink with enhanced recovery precision
Figure BDA0003736677030000052
The digital demodulation is to differentially decode the signals
Figure BDA0003736677030000053
Performing demodulation processing to obtain bit stream information
Figure BDA0003736677030000054
The dequantization process is based on the obtained bit stream information
Figure BDA0003736677030000055
Performing corresponding a-bit dequantization operation, recovering the bit stream information to a real value, and obtaining the recovery information of the downlink CSI
Figure BDA0003736677030000056
The enhanced CSI network Enh-CsiNEt further comprises:
an input layer, a hidden layer containing a Leaky ReLU activation function, and an output layer containing a linear activation function. The number of nodes of the input layer, the hidden layer and the output layer is respectively 2M, nM and 2M, and n represents a hidden layer node coefficient determined according to engineering presetting;
constructing a training data set
Figure BDA0003736677030000057
Training the enhanced CSI network Enh-CsiNet to obtain network parameters of the enhanced CSI network Enh-CsiNet;
the training label
Figure BDA0003736677030000058
The downlink CSI vector h is obtained by real-valued processing, that is:
h label ＝[Re(h 1 ),Im(h 1 ),…,Re(h M ),Im(h M )] T
the training loss function of the enhanced CSI network Enh-CsiNet adopts a mean square error loss function.
In online operation, recovering information of downlink CSI
Figure BDA0003736677030000059
Inputting the enhanced CSI network Enh-CsiNet to finally obtain the downlink with enhanced recovery precision
Figure BDA00037366770300000510
The invention provides a differential coding assisted machine learning channel state information feedback device, a user side device comprises:
the modulation module is used for obtaining a modulated signal S according to the design of the S1;
the differential coding module is used for obtaining a differential coding signal z according to the design in the S1;
the differential coding module is connected behind the modulation signal acquisition module.
The base station side device includes:
a differential decoding module for obtaining differential decoding signal according to the design of S2
Figure BDA00037366770300000511
A precision enhancing module for obtaining the downlink with enhanced recovery precision according to the design of S3
Figure BDA00037366770300000512
The modulation module includes:
quantization unit for real value CSI information
Figure BDA00037366770300000513
A-bit quantization is carried out to obtain real and imaginary bit stream information
Figure BDA00037366770300000514
A modulation unit for converting the bit stream information
Figure BDA00037366770300000515
Performing digital modulation to form modulated signal
Figure BDA00037366770300000516
The accuracy enhancement module comprises:
a demodulation unit for differentially decoding the signals
Figure BDA00037366770300000517
Performing demodulation processing to obtain bit stream information
Figure BDA00037366770300000518
A dequantizing unit for dequantizing the bit stream information based on the obtained bit stream information
Figure BDA0003736677030000061
Performing corresponding a-bit dequantization operation to obtain recovery information of downlink CSI
Figure BDA0003736677030000062
A network enhancing unit for recovering the downlink CSI
Figure BDA0003736677030000063
Inputting the enhanced CSI network Enh-CsiNet to obtain the downlink with enhanced recovery precision
Figure BDA0003736677030000064
Example 1:
in step a 1), a specific embodiment of differentially encoding the modulated signal s to obtain a differentially encoded signal z is as follows:
taking QPSK modulation (modulation order γ = 2), 4-bit quantization (a = 4) as an example, assuming that the downlink CSI vector h length M =2,
the length N =8,s of the modulated signal s is:
Figure BDA0003736677030000065
according to differential coding rules
Figure BDA0003736677030000066
The differentially encoded signal z can be calculated as:
Figure BDA0003736677030000067
example 2:
in step a 2), the base station terminal obtains the difference of the received signal r by adopting a character-by-character differential decoding modeSub-decoding signal
Figure BDA0003736677030000068
One specific example of (a) is as follows:
assuming a received signal
Figure BDA0003736677030000069
According to the word-by-word Fu Chafen coding rule
Figure BDA00037366770300000610
Carrying out differential decoding on the received signal r to obtain a differential decoded signal
Figure BDA00037366770300000611
To be provided with
Figure BDA00037366770300000612
The first element of (1)
Figure BDA00037366770300000613
For example, the specific solving process is as follows:
constellation set
Figure BDA00037366770300000614
In common comprises
Figure BDA00037366770300000615
Four constellation points, first the first one
Figure BDA00037366770300000616
Substituted into Re [ s ] t ·r i * r i-1 ]In (1);
then, sequentially substituting the second to the fourth constellation points to obtain results of 0,0-1;
the result obtained when substituting the first constellation point is maximum, then the result is obtained
Figure BDA00037366770300000617
By the same method for the residue in rThe remaining elements are solved in sequence, and then the signal is differentially decoded
Figure BDA00037366770300000618
The final can be expressed as:
Figure BDA0003736677030000071
example 3:
in the step a 3), recovering the downlink CSI
Figure BDA0003736677030000072
Inputting the enhanced CSI network Enh-CsiNet to obtain downlink with enhanced recovery precision
Figure BDA0003736677030000073
One specific example of (a) is as follows:
assuming downlink CSI recovery information
Figure BDA0003736677030000074
Comprises the following steps:
Figure BDA0003736677030000075
recovering information of downlink CSI
Figure BDA0003736677030000076
Inputting the enhanced CSI network Enh-CsiNet to obtain the real value downlink with enhanced precision
Figure BDA0003736677030000077
Comprises the following steps:
Figure BDA0003736677030000078
it is to be understood that the embodiments described herein are for the purpose of assisting the reader in understanding the manner of practicing the invention and are not to be construed as limiting the scope of the invention to such particular statements and embodiments. Those skilled in the art can make various other specific changes and combinations based on the teachings of the present invention without departing from the spirit of the invention, and these changes and combinations are within the scope of the invention.

Claims (6)

1. The differential coding assisted machine learning channel state information feedback method is characterized by comprising the following steps:
user side processing:
s1, carrying out downlink CSI vector
Figure FDA0003736677020000011
Carrying out real-valued, quantitative and digital modulation to obtain a modulated signal s, carrying out differential coding to obtain a differential coded signal z, carrying out OFDM modulation on the differential coded signal z, and feeding back the differential coded signal z to a base station;
the real-valued transformation is to transform the downlink CSI vector h into real-valued CSI information
Figure FDA0003736677020000012
The quantization is to the real value CSI information
Figure FDA0003736677020000013
A-bit quantization is carried out to obtain real and imaginary bit stream information
Figure FDA0003736677020000014
The digital modulation is to convert bit stream information
Figure FDA0003736677020000015
Performing digital modulation to form modulated signal
Figure FDA0003736677020000016
Wherein the content of the first and second substances,
Figure FDA0003736677020000017
is a modulation order and is a positive integer;
the differential coding is to modulate a signal s = [ s ] 1 ,s 2 ,...,s N ] T According to the rule
Figure FDA0003736677020000018
Carrying out differential coding to obtain a differential coded signal z = [ z ] 1 ,z 2 ,...z N ] T ；
Wherein N =2 aM/gamma, and the differential encoding signal z is a discard z 0 Differential coded signal after = 1;
base station end processing:
s2, the base station side carries out OFDM demodulation to obtain a received signal
Figure FDA0003736677020000019
And a differential decoding signal is obtained by adopting a character-by-character differential decoding mode for the received signal r
Figure FDA00037366770200000110
The differential decoding is realized by solving
Figure FDA00037366770200000111
Obtaining;
wherein the content of the first and second substances,
Figure FDA00037366770200000112
represents the ith decision value; r is i 、r i-1 Respectively represent the i-th element and the i-1-th element in r; s t Represented in a constellation set
Figure FDA00037366770200000113
The trial constellation points in (1) take values, superscript denotes taking conjugate operation, re [ ·]Representing the real part, max [. Cndot]Representing a maximum value operation;
s3, will be poorSub-decoding signal
Figure FDA00037366770200000114
Performing digital demodulation and dequantization to obtain downlink CSI recovery information
Figure FDA00037366770200000115
And constructing an enhanced CSI network Enh-CsiNet to recover information of downlink CSI
Figure FDA00037366770200000116
Enhancing to obtain downlink CSI with enhanced recovery precision
Figure FDA00037366770200000117
The digital demodulation is to differentially decode the signals
Figure FDA00037366770200000118
Performing demodulation processing to obtain bit stream information
Figure FDA00037366770200000119
The dequantization process is based on the obtained bit stream information
Figure FDA00037366770200000120
Corresponding a-bit dequantization operation is carried out to obtain the recovery information of the downlink CSI
Figure FDA00037366770200000121
2. The feedback method of machine learning channel state information assisted by differential coding according to claim 1, wherein the enhanced CSI network Enh-CsiNEt of step S3 further comprises:
an input layer, a hidden layer containing a Leaky Relu activation function, and an output layer containing a linear activation function; the number of nodes of the input layer, the hidden layer and the output layer is respectively 2M, nM and 2M, and n represents a hidden layer node coefficient determined according to engineering presetting;
constructing a training data set
Figure FDA0003736677020000021
Training the enhanced CSI network Enh-CsiNet to obtain network parameters of the enhanced CSI network Enh-CsiNet;
the training label
Figure FDA0003736677020000022
The downlink CSI vector h is obtained by real-valued processing, that is:
h label ＝[Re(h 1 ),Im(h 1 ),…,Re(h M ),Im(h M )] T
wherein Im (·) represents taking the imaginary part, [ ·] T Representing a transposition operation;
when in online operation, the recovery information of the downlink CSI is transmitted
Figure FDA0003736677020000023
Inputting the enhanced CSI network Enh-CsiNet to obtain the downlink CSI with enhanced recovery precision
Figure FDA0003736677020000024
3. A differential coding assisted machine learning channel state information feedback device for implementing the differential coding assisted machine learning channel state information feedback method of claim 1 or 2;
the user side device includes:
a modulation module configured to obtain a modulated signal S according to the design of S1 in claim 1;
differential coding module, which obtains a differential coded signal z according to the design described in S1 in claim 1.
4. The differential encoding assisted machine learning channel state information feedback apparatus of claim 3, wherein the modulation module comprises:
quantization unit for real value CSI information
Figure FDA0003736677020000025
A-bit quantization is carried out to obtain real and imaginary bit stream information
Figure FDA0003736677020000026
A modulation unit for converting the bit stream information
Figure FDA0003736677020000027
Performing digital modulation to form modulated signal
Figure FDA0003736677020000028
5. The differential coding assisted machine learning channel state information feedback device of claim 3, wherein the base station side device comprises:
differential decoding module for obtaining a differential decoded signal according to the design of S2 in claim 1
Figure FDA0003736677020000029
The accuracy enhancement module obtains the downlink CSI with enhanced recovery accuracy according to the design of S3 in claim 1
Figure FDA00037366770200000210
6. The differential coding assisted machine learning channel state information feedback device of claim 5, wherein the precision enhancement module comprises:
a demodulation unit for differentially decoding the signals
Figure FDA00037366770200000211
Performing demodulation processing to obtain bit stream information
Figure FDA00037366770200000212
A dequantizing unit for dequantizing the bit stream information based on the obtained bit stream information
Figure FDA00037366770200000213
Corresponding a-bit dequantization operation is carried out to obtain the recovery information of the downlink CSI
Figure FDA00037366770200000214
A network enhancing unit for recovering the downlink CSI
Figure FDA00037366770200000215
Inputting the enhanced CSI network Enh-CsiNet to obtain the downlink CSI with enhanced recovery precision
Figure FDA0003736677020000031
CN202210805014.5A 2022-07-08 2022-07-08 Differential coding assisted machine learning channel state information feedback method and device Pending CN115189742A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210805014.5A CN115189742A (en) 2022-07-08 2022-07-08 Differential coding assisted machine learning channel state information feedback method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210805014.5A CN115189742A (en) 2022-07-08 2022-07-08 Differential coding assisted machine learning channel state information feedback method and device

Publications (1)

Publication Number Publication Date
CN115189742A true CN115189742A (en) 2022-10-14

Family

ID=83517147

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210805014.5A Pending CN115189742A (en) 2022-07-08 2022-07-08 Differential coding assisted machine learning channel state information feedback method and device

Country Status (1)

Country Link
CN (1) CN115189742A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117336125A (en) * 2023-11-28 2024-01-02 西华大学 Decision feedback channel estimation method and device in differential OTFS system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117336125A (en) * 2023-11-28 2024-01-02 西华大学 Decision feedback channel estimation method and device in differential OTFS system

Similar Documents

Publication Publication Date Title
Jiang et al. Deep source-channel coding for sentence semantic transmission with HARQ
Weng et al. Semantic communication systems for speech transmission
Hu et al. Robust semantic communications with masked VQ-VAE enabled codebook
Dai et al. Communication beyond transmitting bits: Semantics-guided source and channel coding
CN109687897B (en) Superposition CSI feedback method based on deep learning large-scale MIMO system
US9020029B2 (en) Arbitrary precision multiple description coding
CN115189742A (en) Differential coding assisted machine learning channel state information feedback method and device
CN112564757A (en) Deep learning 1-bit compression superposition channel state information feedback method
Xu et al. Deep joint source-channel coding for CSI feedback: An end-to-end approach
CN116436567A (en) Semantic communication method based on deep neural network
CN110602017A (en) Non-orthogonal multiple access decoding method
CN113872652A (en) CSI feedback method based on 3D MIMO time-varying system
Xu et al. Transformer empowered CSI feedback for massive MIMO systems
CN113726376B (en) 1bit compression superposition CSI feedback method based on feature extraction and mutual-difference fusion
CN116707712A (en) Superimposed channel state information feedback method for differential modulation auxiliary parallel branch fusion
CN115250217B (en) OFDM system channel estimation method and device assisted by differential coding and neural network
Sheng et al. A multi-task semantic communication system for natural language processing
CN116390134A (en) Semantic communication transmission method based on non-orthogonal multiple access
CN116074414A (en) Wireless communication physical layer structure based on deep learning
Fu et al. Scalable extraction based semantic communication for 6G wireless networks
CN115114928A (en) Interpretable semantic communication system based on feature selection
Wang et al. Transceiver cooperative learning-aided semantic communications against mismatched background knowledge bases
Dong et al. Innovative semantic communication system
Jiang et al. Adaptive semantic video conferencing for ofdm systems
CN114389655A (en) Detection method for incoherent coding of large-scale MIMO system under related channel

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