CN114567527A - Reconfigurable intelligent surface assisted superposition guidance fusion learning channel estimation method - Google Patents

Reconfigurable intelligent surface assisted superposition guidance fusion learning channel estimation method Download PDF

Info

Publication number
CN114567527A
CN114567527A CN202210233217.1A CN202210233217A CN114567527A CN 114567527 A CN114567527 A CN 114567527A CN 202210233217 A CN202210233217 A CN 202210233217A CN 114567527 A CN114567527 A CN 114567527A
Authority
CN
China
Prior art keywords
network
channel estimation
equalization
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.)
Granted
Application number
CN202210233217.1A
Other languages
Chinese (zh)
Other versions
CN114567527B (en
Inventor
卿朝进
王莉
凌国伟
董磊
张岷涛
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 CN202210233217.1A priority Critical patent/CN114567527B/en
Publication of CN114567527A publication Critical patent/CN114567527A/en
Application granted granted Critical
Publication of CN114567527B publication Critical patent/CN114567527B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Power Engineering (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)

Abstract

The invention discloses a reconfigurable intelligent surface assisted superposition guidance fusion learning channel estimation method.A receiver receives a wireless signal of a transmitter adopting a superposition guidance scheme to form a received signal; performing characteristic extraction on the channel state to obtain 'CSI initial characteristic', inputting the 'CSI initial characteristic' into a trained 'enhanced channel estimation network EN-CENet' to obtain 'enhanced CSI'; performing equalization feature extraction on a received signal to obtain an equalization initial feature; eliminating the pilot to obtain pre-equalization data, inputting the pre-equalization data into a trained enhanced symbol detection network EN-SDNet to obtain a detection symbol; and inputting the received signal, the enhanced CSI and the detection symbol into a trained update fusion network Up-FUSNet to obtain an enhanced detection symbol. The invention can improve the channel estimation precision and the symbol detection performance, solves the problem that a propagation path is shielded by using the RIS, and greatly improves the channel estimation performance and the symbol detection performance.

Description

Reconfigurable intelligent surface assisted superposition guidance fusion learning channel estimation method
Technical Field
The invention relates to the technical field of reconfigurable intelligent surface-assisted channel estimation in wireless communication, in particular to a reconfigurable intelligent surface-assisted superposition-guided fusion learning channel estimation method.
Background
The fifth generation (5G) and sixth generation (6G) networks under development will be the cornerstone to achieve the goal of future internet of things connectivity. The internet of things has attracted attention of many people, for example, an intelligent building system connected with an automatic robot can control and manage different intelligent devices through a network, so as to save energy and improve convenience of life of residents. Other applications include smart medicine, smart driving, smart home, and the like. In the internet of things system, channel estimation plays a key role. On one hand, in an internet of things system such as an industrial internet of things, the channel condition is very complex, and a situation that the channel time changes or the blocking probability is increased may be encountered. On the other hand, with good channel estimation, the internet of things device can transmit with affordable transmission power using appropriate modulation and coding methods.
Channel estimation for the system of the internet of things is crucial for efficient receiver operation. Aiming at an internet of things system, a one-dimensional time domain wiener filtering technology is combined with frequency domain maximum likelihood estimation which is simple in calculation, and a pilot frequency-based hybrid channel estimation method is provided. Channel estimation based on least squares and minimum mean square error in an internet of things system has also been studied. An improved computationally efficient linear minimum mean square error estimator for an internet of things system has also been proposed. However, these channel estimation methods inevitably occupy spectrum resources for transmission guidance, which causes huge resource waste for the internet of things system. In contrast, people propose to transmit guidance and data in a superposition manner without additional time-frequency resources, so that the problems of low spectrum efficiency and energy consumption are solved. Therefore, the method using overlay steering is a promising solution.
Although the above-mentioned method of channel estimation by superposition steering solves the spectral efficiency problem, the energy consumption problem of the internet of things system is not well alleviated. The goal of the internet of things system is to extend the battery life of the user to 10 years, which makes it particularly important to solve the energy consumption problem. If the guidance and the data are transmitted separately, the energy consumption is inevitably increased, and the battery life of the terminal of the internet of things is finally reduced. Therefore, in order to reduce the energy consumption of the internet of things system, channel estimation based on superposition guidance needs to be further developed. This prompted us to develop pilot-based superposition channel estimation in this patent.
In addition, due to the complex scene of the internet of things, for example, in the industrial internet of things, the propagation path is blocked, so that the communication of the internet of things is also a common situation. Therefore, how to improve the robustness of the communication link is an urgent problem to be solved. Reconfigurable Intelligent Surface (RIS) offers an attractive option for blocked propagation paths. The RIS is an artificial electromagnetic panel consisting of a large number of low-cost passive scattering elements that control the wireless environment by adjusting the amplitude or phase shift of the reflected signal. Unlike conventional forward amplifying relays, these passive devices consume little energy. Recently, the incorporation of RIS into the internet of things is considered to be a revolutionary means of transforming any passive wireless communication environment into an active wireless communication environment. In addition, the RIS also improves system throughput by at least 40% and improves system coverage by 1/3. Therefore, the RIS is deployed in the Internet of things system to be an ideal way for solving the problem of the blockage of the propagation path of the Internet of things. However, to our knowledge, this solution has not been well studied in the existing work.
Disclosure of Invention
The invention aims to provide a reconfigurable intelligent surface assisted superposition guidance fusion learning channel estimation method, which considers the energy service life problem and the shielding condition of a propagation path compared with the existing channel estimation method, and effectively improves the system performance by means of a channel estimation network, a symbol detection network and a fusion network.
The technical scheme of the invention is as follows:
a reconfigurable intelligent surface assisted superposition guidance fusion learning channel estimation method comprises the following steps:
s1, the receiver receives the wireless signal of the transmitter adopting the superposition guidance scheme to form a received signal y with the length of N;
the wireless signal, undergoes an RIS reflection;
s2, according to the received signal y, performs feature extraction on the Channel State Information (CSI) to obtain the 'CSI initial feature' with the length of N "
Figure BDA0003540651670000021
The CSI characteristic extraction method comprises the traditional LS and MMSE linear channel estimation and nonlinear channel estimation methods such as maximum likelihood based on Bayes, Markov Monte Carlo and nonlinear filter;
s3 will "CSI initial characteristics"
Figure BDA0003540651670000022
Inputting the signal into a trained enhanced channel estimation network EN-CENet to obtain enhanced CSI with the length of N "
Figure BDA0003540651670000023
S4 according to "enhanced CSI"
Figure BDA0003540651670000024
Carrying out equalization feature extraction on the received signal y to obtain an equalization initial feature with the length of N "
Figure BDA0003540651670000025
The equalization characteristic extraction method comprises the steps of traditional Zero-Forcing (ZF) equalization, MMSE equalization, LMS equalization, RLS equalization and other characteristic extraction;
s5 based on "equalizing initial features"
Figure BDA0003540651670000026
Eliminating "pilot" xpTo obtain "Pre-equalization data"
Figure BDA0003540651670000027
S6 will "Pre-equalization data"
Figure BDA0003540651670000028
Inputting the signal into a trained enhanced symbol detection network EN-Net to obtain a detection symbol with the length of N
Figure BDA0003540651670000029
S7 converts the received signal y into "enhanced CSI"
Figure BDA00035406516700000210
And detecting the symbol
Figure BDA00035406516700000211
Inputting the data into a trained 'update fusion network Up-FUSNet' to obtain an 'enhanced detection symbol' with the length of N "
Figure BDA00035406516700000212
According to some embodiments of the invention, step S1 further comprises:
the S11 receiver receives the wireless signal generated by the transmitter via RIS using the superposition steering scheme, forming a received signal y of length N, which is expressed as follows:
y＝h⊙x+n；
wherein n represents a zero mean variance of
Figure BDA0003540651670000031
A circularly symmetric complex Gaussian noise of x denotes a transmission signal, < > denotes a Hadamard product, and h denotes a complex channel frequency response formed by the RIS, is formed by the direct link hTRAnd a reflective link HTRRComposition, expressed as follows:
h＝hTR+HTRRφ；
wherein phi is [ phi ]1,φ2,…,φM]Represents the phase shift vector, as follows:
Figure BDA0003540651670000032
wherein ,
Figure BDA0003540651670000033
represents the phase shift of the mth RIS subsurface, total M RIS subsurface, betamRepresenting the amplitude of the RIS surface, designed to satisfy beta according to engineering experiencem＝1；
Wherein the reflective link HTRR＝[hTRR,1,hTRR,2,…,hTRR,m,…,hTRR,M]Reflection link h of mth RIS subsurfaceTRR,mExpressed as follows:
hTRR,m＝hTR,m⊙hRR,m；
wherein ,hTR,mDenotes the channel frequency response, h, of the mth RIS sub-surface transmitter to the RIS linkRR,mRepresenting the RIS to receiver link channel frequency response for the mth RIS subsurface.
According to some embodiments of the invention, step S2 further comprises:
s21 constructs a superimposed transmit signal x, represented as follows:
Figure BDA0003540651670000034
where ρ represents a power scaling factor, E represents a transmit power, xPDenotes guide, xdRepresenting modulation symbols, transmitted signal x being steered xPAnd modulation symbol xdForming by superposition;
the S22 receiver receives the RIS-generated wireless signal to form a received signal y, which is expressed as follows:
y＝h⊙x+n；
s23, extracting the characteristics of Channel State Information (CSI) to obtain the length N 'CSI initial characteristics'
Figure BDA0003540651670000035
The CSI characteristic extraction method comprises traditional LS and MMSE linear channel estimation and nonlinear channel estimation methods such as maximum likelihood based on Bayes, Markov Monte Carlo and nonlinear filter, and the LS estimation is taken as an example and expressed as follows:
Figure BDA0003540651670000036
wherein ,(·)TIndicating transposition.
According to some embodiments of the invention, step S3 further comprises:
1 single hidden layer, 1 input layer and 1 output layer, wherein the number of neurons in each layer is 2N;
carrying out normalization processing on hidden layer output, limiting an output value to a range [0,1], and then outputting an overactivation function to the normalized hidden layer;
the hidden layer activation function is:
Figure BDA0003540651670000041
the input layer and the output layer adopt linear activation functions;
setting the loss function as a mean square error function;
using training sets
Figure BDA0003540651670000042
Training the network, and storing a network model and parameters thereof after error convergence;
the training input of the network is
Figure BDA0003540651670000043
The training label is
Figure BDA0003540651670000044
The training label is obtained by measuring according to the actual scene, modeling a channel model and finally outputting 'enhanced CSI' by a network "
Figure BDA0003540651670000045
According to some embodiments of the invention, step S4 further comprises:
s41 based on "enhanced CSI"
Figure BDA0003540651670000046
Carrying out equalization feature extraction on the received signal y to obtain an equalization initial feature with the length of N "
Figure BDA0003540651670000047
The equalization feature extraction method comprises the traditional Zero-Forcing (ZF) equalization, MMSE equalization, LMS equalization, RLS equalization and other feature extraction, and the following takes ZF equalization as an example and shows the following steps:
Figure BDA0003540651670000048
wherein ,GEQAn equalization matrix is represented as follows:
Figure BDA0003540651670000049
wherein the equalization matrix GEQIts element is read "enhanced CSI"
Figure BDA00035406516700000410
Obtaining the value of (A);
wherein, the 'enhanced CSI'
Figure BDA00035406516700000411
The vector form of (a) is:
Figure BDA00035406516700000412
according to some embodiments of the invention, step S5 further comprises:
s51 will "balance initial features"
Figure BDA00035406516700000413
Eliminating "pilot" xpTo obtain "Pre-equalization data"
Figure BDA00035406516700000414
Is represented as follows:
Figure BDA00035406516700000415
according to some embodiments of the invention, step S6 further comprises:
p (p is more than or equal to 2) hidden layers, 1 input layer and 1 output layer;
the number of neurons of the input layer and the output layer is set to be 2N, and the number of neurons of each hidden layer is set to be qN;
the activation function of the first hidden layer is set as LeakyReLU, the activation functions of the rest n-1 hidden layers are set as ReLU, and the input layer and the output layer both adopt linear activation functions;
collecting training sets
Figure BDA0003540651670000051
Training the network, and storing a network model and parameters thereof after error convergence;
the number n (n is more than or equal to 2) of the hidden layers and the number qN of the nodes of the hidden layers are subjected to parameter tuning and setting according to engineering experience;
the training input of the network is
Figure BDA0003540651670000052
Training label xdTraining labels are modulation symbols and finally the output of the network is a detection symbol
Figure BDA0003540651670000053
According to some embodiments of the invention, step S7 further comprises:
the Up-FUSNet updating fusion network comprises a single hidden layer, 1 input layer and 1 output layer;
the number of neuron in hidden layer of Up-FUSNet of' update fusion network is u1N, the number of neurons in the input layer and the output layer is 2N;
the hidden layer adopts an activation function ReLU, and the input layer and the output layer both adopt linear activation functions;
setting a loss function of the network as a mean square error function;
collecting training sets
Figure BDA0003540651670000054
Training the network, and storing a network model and parameters thereof after error convergence;
the number u of hidden layer neurons1N, carrying out parameter tuning and setting according to engineering experience;
the training input of the network is
Figure BDA0003540651670000055
From the received signal y, "enhanced CSI"
Figure BDA0003540651670000056
And detecting the symbol
Figure BDA0003540651670000057
Formed by the training labels being modulation symbols xdThe output of the final network is "enhanced detection symbols"
Figure BDA0003540651670000058
The invention constructs three neural networks: the channel estimation network, the symbol detection network and the updating fusion network are enhanced, the Bit Error Rate (BER) is reduced, and simulation results show that the Bit Error Rate of the updating fusion network is up to the Bit Error Rate of an ideal channel after MMSE equalization.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
fig. 2 is a schematic diagram of a CSI feature extraction process.
Detailed Description
The present invention is described in detail below with reference to the following embodiments and the attached drawings, but it should be understood that the embodiments and the attached drawings are only used for the illustrative description of the present invention and do not limit the protection scope of the present invention in any way. All reasonable variations and combinations that fall within the spirit of the invention are intended to be within the scope of the invention.
According to the technical scheme of the invention, the channel estimation method for the reconfigurable intelligent surface assisted superposition guidance fusion learning comprises the flow shown in the attached figure 1, and specifically comprises the following steps:
setting: the number of subcarriers N is 6, the number of multipath L is 4, the number of RIS surfaces M is 4, and rho is 0.15;
in step S2, according to the received signal y, feature extraction is performed on the Channel State Information (CSI), as shown in the flow chart of fig. 2, to obtain "CSI initial feature" with length N "
Figure BDA0003540651670000061
Figure BDA0003540651670000062
In step S3, the "CSI initial characteristics"
Figure BDA0003540651670000063
Inputting the signal into a trained enhanced channel estimation network EN-CENet to obtain enhanced CSI with the length of N "
Figure BDA0003540651670000064
Figure BDA0003540651670000065
According to the reference original channel value h, the following is:
Figure BDA0003540651670000066
the normalized mean square error NMSE at a signal-to-noise ratio SNR of 0,5,10, …,25dB can be found as follows:
NMSE_LS＝[1.48,1.38,1.29,1.28,1.26,1.26]；
NMSE_Net＝[0.04,0.03,0.02,0.016,0.014,0.013]；
it can be seen that the NMSE value is reduced and the channel estimation performance is improved after the channel estimation model is used;
in step S5, "balance initial characteristics"
Figure BDA0003540651670000067
Eliminating "pilot" xpTo obtain "Pre-equalization data"
Figure BDA0003540651670000068
Figure BDA0003540651670000071
In step S7, the received signal y is "enhanced CSI"
Figure BDA0003540651670000072
And detecting the symbol
Figure BDA0003540651670000073
Inputting the data into a trained 'update fusion network Up-FUSNet' to obtain an 'enhanced detection symbol' with the length of N "
Figure BDA0003540651670000074
Figure BDA0003540651670000075
Modulation symbol x according to referencedThe following are:
Figure BDA0003540651670000076
the BER at SNR of 0,5,10, …,25dB can be determined as follows:
BER＝[0.93,0.82,0.67,0.66,0.63,0.63]；
BER_Net＝[0.204,0.086,0.014,0.067,0.083,0]；
it can be seen that after the updated converged network model is used, the bit error rate BER is reduced, and the symbol detection performance is improved.
The above examples are merely preferred embodiments of the present invention, and the scope of the present invention is not limited to the above examples. All technical schemes belonging to the idea of the invention belong to the protection scope of the invention. It should be noted that modifications and adaptations to those skilled in the art without departing from the principles of the present invention should also be considered as within the scope of the present invention.

Claims (8)

1. A reconfigurable intelligent surface assisted superposition guidance fusion learning channel estimation method is characterized by comprising the following steps:
s1, the receiver receives the wireless signal of the transmitter adopting the superposition guidance scheme to form a received signal y with the length of N;
the wireless signal, undergoes an RIS reflection;
s2, according to the received signal y, extracting the characteristics of the channel state information to obtain the 'CSI initial characteristics' with the length of N "
Figure FDA0003540651660000011
S3 will "CSI initial characteristics"
Figure FDA0003540651660000012
Inputting the signal into a trained enhanced channel estimation network EN-CENet to obtain enhanced CSI with the length of N "
Figure FDA0003540651660000013
S4 based on "enhanced CSI"
Figure FDA0003540651660000014
Carrying out equalization feature extraction on the received signal y to obtain an equalization initial feature with the length of N "
Figure FDA0003540651660000015
S5 based on "Balancing initial features"
Figure FDA0003540651660000016
Eliminating "pilot" xpTo obtain "Pre-equalization data"
Figure FDA0003540651660000017
S6 will "Pre-equalization data"
Figure FDA0003540651660000018
Inputting the signal into a trained enhanced symbol detection network EN-Net to obtain a detection symbol with the length of N
Figure FDA0003540651660000019
S7 receiving signal y, enhanced CSI "
Figure FDA00035406516600000110
And detecting the symbol
Figure FDA00035406516600000111
Inputting the data into a trained 'update fusion network Up-FUSNet' to obtain an 'enhanced detection symbol' with the length of N "
Figure FDA00035406516600000112
2. The method for channel estimation through guiding fusion learning of reconfigurable intelligent surface-assisted superposition according to claim 1, wherein the step S1 includes the following sub-steps:
the S11 receiver receives the radio signal generated by the transmitter through RIS using the superposition steering scheme to form a received signal y with a length N, which is expressed as follows:
y＝h⊙x+n；
wherein n represents a zero mean variance of
Figure FDA00035406516600000113
A circularly symmetric complex Gaussian noise of x denotes a transmission signal, < > denotes a Hadamard product, and h denotes a complex channel frequency response formed by the RIS, is formed by the direct link hTRAnd a reflective link HTRRComposition, expressed as follows:
h＝hTR+HTRRφ；
wherein phi is [ phi ]1,φ2,…,φM]Representing the phase shift vector as follows:
Figure FDA00035406516600000114
wherein ,
Figure FDA00035406516600000115
represents the phase shift of the mth RIS subsurface, total M RIS subsurface, betamRepresenting the amplitude of the RIS surface, designed to satisfy beta according to engineering experiencem＝1；
Wherein the reflective link HTRR＝[hTRR,1,hTRR,2,…,hTRR,m,…,hTRR,M]Reflective link h of mth RIS subsurfaceTRR,mExpressed as follows:
hTRR,m＝hTR,m⊙hRR,m；
wherein ,hTR,mRepresents the channel frequency response, h, of the mth RIS subsurface transmitter to RIS linkRR,mRepresenting the RIS to receiver link channel frequency response for the mth RIS subsurface.
3. The method for estimating the channel through the reconfigurable intelligent surface-assisted superposition-guided fusion learning as claimed in claim 1, wherein the Channel State Information (CSI) is subjected to feature extraction in step S2 to obtain a 'CSI initial feature' with the length of N "
Figure FDA0003540651660000021
The CSI characteristic extraction method comprises the traditional LS and MMSE linear channel estimation and nonlinear channel estimation methods based on Bayes, Markov Monte Carlo and the maximum likelihood of a nonlinear filter.
4. The method for channel estimation through guiding fusion learning of reconfigurable intelligent surface-assisted superposition according to claim 1, wherein the step S3 includes the following sub-steps:
1 single hidden layer, 1 input layer and 1 output layer, wherein the number of neurons in each layer is 2N;
carrying out normalization processing on hidden layer output, limiting an output value to a range [0,1], and then outputting an overactivation function to the normalized hidden layer;
the hidden layer activation function is:
Figure FDA0003540651660000022
the input layer and the output layer adopt linear activation functions;
setting the loss function as a mean square error function;
using training sets
Figure FDA0003540651660000023
Training the network, and storing a network model and parameters thereof after error convergence;
the training input of the network is
Figure FDA0003540651660000024
The training labels are
Figure FDA0003540651660000025
The training label is obtained by measuring according to the actual scene, modeling a channel model and finally outputting 'enhanced CSI' by a network "
Figure FDA0003540651660000026
5. The method for channel estimation through guiding fusion learning of reconfigurable intelligent surface-assisted superposition according to claim 1, wherein the step S4 includes the following sub-steps:
s41 based on "enhanced CSI"
Figure FDA0003540651660000027
Carrying out equalization feature extraction on the received signal y to obtain an equalization initial feature with the length of N "
Figure FDA0003540651660000028
The equalization characteristic extraction method comprises the traditional zero forcing ZF equalization, MMSE equalization, LMS equalization, RLS equalization and other characteristic extraction.
6. The method for channel estimation through guiding fusion learning of reconfigurable intelligent surface-assisted superposition according to claim 1, wherein the step S5 includes the following sub-steps:
will balance the initial characteristics "
Figure FDA0003540651660000031
Eliminating "pilot" xpTo obtain "Pre-equalization data"
Figure FDA0003540651660000032
Is represented as follows:
Figure FDA0003540651660000033
7. the method for channel estimation through guiding fusion learning of reconfigurable intelligent surface-assisted superposition according to claim 1, wherein the step S6 includes the following sub-steps:
p (p is more than or equal to 2) hidden layers, 1 input layer and 1 output layer;
the number of neurons of the input layer and the output layer is set to be 2N, and the number of neurons of each hidden layer is set to be qN;
the activation function of the first hidden layer is set as LeakyReLU, the activation functions of the rest n-1 hidden layers are set as ReLU, and the input layer and the output layer both adopt linear activation functions;
collecting training sets
Figure FDA0003540651660000034
Training the network, and storing a network model and parameters thereof after error convergence;
the number n (n is more than or equal to 2) of the hidden layers and the number qN of the nodes of the hidden layers are subjected to parameter tuning and setting according to engineering experience;
the training input of the network is
Figure FDA0003540651660000035
Training label xdTraining labels are modulation symbols and finally the output of the network is a detection symbol
Figure FDA0003540651660000036
8. The method for channel estimation of reconfigurable intelligent surface-assisted superposition-guided fusion learning according to claim 1, wherein the step S7 includes the following sub-steps:
the Up-FUSNet updating fusion network comprises a single hidden layer, 1 input layer and 1 output layer;
the number of neuron in hidden layer of Up-FUSNet of' update fusion network is u1N, number of input and output layer neuronsAre all 2N;
the hidden layer adopts an activation function ReLU, and the input layer and the output layer both adopt linear activation functions;
setting a loss function of the network as a mean square error function;
collecting training sets
Figure FDA0003540651660000037
Training the network, and storing a network model and parameters thereof after error convergence;
the number u of hidden layer neurons1N, carrying out parameter tuning and setting according to engineering experience;
the training input of the network is
Figure FDA0003540651660000038
From the received signal y, "enhanced CSI"
Figure FDA0003540651660000039
And detecting the symbol
Figure FDA00035406516600000310
Formed by the training labels being modulation symbols xdThe output of the final network is "enhanced detection symbols"
Figure FDA00035406516600000311
CN202210233217.1A 2022-03-10 2022-03-10 Reconfigurable intelligent surface auxiliary superposition guide fusion learning channel estimation method Active CN114567527B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210233217.1A CN114567527B (en) 2022-03-10 2022-03-10 Reconfigurable intelligent surface auxiliary superposition guide fusion learning channel estimation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210233217.1A CN114567527B (en) 2022-03-10 2022-03-10 Reconfigurable intelligent surface auxiliary superposition guide fusion learning channel estimation method

Publications (2)

Publication Number Publication Date
CN114567527A true CN114567527A (en) 2022-05-31
CN114567527B CN114567527B (en) 2023-04-28

Family

ID=81717300

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210233217.1A Active CN114567527B (en) 2022-03-10 2022-03-10 Reconfigurable intelligent surface auxiliary superposition guide fusion learning channel estimation method

Country Status (1)

Country Link
CN (1) CN114567527B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115250217A (en) * 2022-07-22 2022-10-28 西华大学 Differential coding and neural network assisted OFDM system channel estimation method and device

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014109613A1 (en) * 2013-01-14 2014-07-17 엘지전자 주식회사 Method and apparatus for transmitting/receiving channel state information in wireless communication system
CN110706709A (en) * 2019-08-30 2020-01-17 广东工业大学 Multi-channel convolution aliasing voice channel estimation algorithm combined with video signal
CN113381953A (en) * 2021-06-11 2021-09-10 西华大学 Channel estimation method of extreme learning machine based on reconfigurable intelligent surface assistance
US20210314036A1 (en) * 2020-04-03 2021-10-07 Samsung Electronics Co., Ltd. Rank indicator (ri) and channel quality indicator (cqi) estimation using a multi-layer perceptron (mlp)
WO2022007417A1 (en) * 2020-07-10 2022-01-13 Huawei Technologies Co., Ltd. Systems and methods using configurable surfaces for wireless communication
CN114124623A (en) * 2021-10-28 2022-03-01 中国信息通信研究院 Wireless communication channel estimation method and device

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014109613A1 (en) * 2013-01-14 2014-07-17 엘지전자 주식회사 Method and apparatus for transmitting/receiving channel state information in wireless communication system
CN110706709A (en) * 2019-08-30 2020-01-17 广东工业大学 Multi-channel convolution aliasing voice channel estimation algorithm combined with video signal
US20210314036A1 (en) * 2020-04-03 2021-10-07 Samsung Electronics Co., Ltd. Rank indicator (ri) and channel quality indicator (cqi) estimation using a multi-layer perceptron (mlp)
WO2022007417A1 (en) * 2020-07-10 2022-01-13 Huawei Technologies Co., Ltd. Systems and methods using configurable surfaces for wireless communication
CN113381953A (en) * 2021-06-11 2021-09-10 西华大学 Channel estimation method of extreme learning machine based on reconfigurable intelligent surface assistance
CN114124623A (en) * 2021-10-28 2022-03-01 中国信息通信研究院 Wireless communication channel estimation method and device

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
KANGDA ZHI等: "Statistical CSI-Based Design for Reconfigurable Intelligent Surface-Aided Massive MIMO Systems With Direct Links" *
卿朝进等: "Enhanced ELM Based Channel Estimation for RIS-Assisted OFDM Systems With Insufficient CP and Imperfect Hardware" *
程执天: "5GNR大规模天线系统中信道状态信息增强技术研究与性能评估" *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115250217A (en) * 2022-07-22 2022-10-28 西华大学 Differential coding and neural network assisted OFDM system channel estimation method and device
CN115250217B (en) * 2022-07-22 2023-05-19 西华大学 OFDM system channel estimation method and device assisted by differential coding and neural network

Also Published As

Publication number Publication date
CN114567527B (en) 2023-04-28

Similar Documents

Publication Publication Date Title
CN110460402B (en) End-to-end communication system establishing method based on deep learning
CN102484624B (en) Symbol estimation methods and apparatuses
Wu et al. Deep learning-based autoencoder for m-user wireless interference channel physical layer design
CN114051701B (en) Apparatus and method for machine learning assisted precoding
CN112583458B (en) MIMO end-to-end transmission system based on deep learning and wireless transformation network
Shah et al. Fractional order constant modulus blind algorithms with application to channel equalisation
CN114567527B (en) Reconfigurable intelligent surface auxiliary superposition guide fusion learning channel estimation method
CN112291005A (en) Bi-LSTM neural network-based receiving end signal detection method
Hua et al. Signal detection in uplink pilot-assisted multi-user MIMO systems with deep learning
CN114499601B (en) Large-scale MIMO signal detection method based on deep learning
CN113630130A (en) End-to-end digital communication demodulation method
Wang et al. Online LSTM-based channel estimation for HF MIMO SC-FDE system
Zhang et al. Machine learning for joint channel equalization and signal detection
Seyman et al. MIMO-OFDM channel estimation using ANFIS
CN101651643A (en) Blind equalization method for wavelet neural network based on space diversity
Yıldırım et al. Deep receiver design for multi-carrier waveforms using cnns
Zhao et al. Model-Driven Based Deep Unfolding Equalizer for Underwater Acoustic OFDM Communications
Myburgh et al. Low complexity MLSE equalization in highly dispersive Rayleigh fading channels
Hanoon et al. Performance Evaluation of SISO-OFDM Channel Equalization Utilizing Deep Learning
Bhuyan et al. Nonlinear mobile link adaptation using modified FLNN and channel sounder arrangement
Ajayi et al. Low-Complexity Neural Networks for Denoising Imperfect CSI in Physical Layer Security
Reddy et al. Hybrid optimization‐based deep neuro‐fuzzy network for designing m‐user multiple‐input multiple‐output interference channel
CN113300788B (en) Blind receiver method and device based on Capsule network
Muthulakshmi et al. Signal demodulation without channel equalizer using machine learning techniques
Ramasamy et al. Evolutionary gravitational Neocognitron neural network based differential spatial modulation detection scheme for uplink multiple user huge MIMO systems

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20220531

Assignee: Chengdu Suyouyun Information Technology Co.,Ltd.

Assignor: XIHUA University

Contract record no.: X2023510000030

Denomination of invention: A reconfigurable intelligent surface assisted overlay guidance fusion learning channel estimation method

Granted publication date: 20230428

License type: Common License

Record date: 20231201

EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20220531

Assignee: Chengdu Yingling Feifan Technology Co.,Ltd.

Assignor: XIHUA University

Contract record no.: X2023510000032

Denomination of invention: A reconfigurable intelligent surface assisted overlay guidance fusion learning channel estimation method

Granted publication date: 20230428

License type: Common License

Record date: 20231212

EE01 Entry into force of recordation of patent licensing contract