CN111884976B - Channel interpolation method based on neural network - Google Patents

Channel interpolation method based on neural network Download PDF

Info

Publication number
CN111884976B
CN111884976B CN202010705053.9A CN202010705053A CN111884976B CN 111884976 B CN111884976 B CN 111884976B CN 202010705053 A CN202010705053 A CN 202010705053A CN 111884976 B CN111884976 B CN 111884976B
Authority
CN
China
Prior art keywords
neural network
training
channel
frequency domain
vector
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.)
Active
Application number
CN202010705053.9A
Other languages
Chinese (zh)
Other versions
CN111884976A (en
Inventor
李靖
韩竞宇
葛建华
任德锋
李慧芳
高明
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.)
Xidian University
Original Assignee
Xidian 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 Xidian University filed Critical Xidian University
Priority to CN202010705053.9A priority Critical patent/CN111884976B/en
Publication of CN111884976A publication Critical patent/CN111884976A/en
Application granted granted Critical
Publication of CN111884976B publication Critical patent/CN111884976B/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
    • H04L27/00—Modulated-carrier systems
    • H04L27/26—Systems using multi-frequency codes
    • H04L27/2601—Multicarrier modulation systems
    • H04L27/2647—Arrangements specific to the receiver only
    • H04L27/2655—Synchronisation arrangements
    • H04L27/2689—Link with other circuits, i.e. special connections between synchronisation arrangements and other circuits for achieving synchronisation
    • H04L27/2695—Link with other circuits, i.e. special connections between synchronisation arrangements and other circuits for achieving synchronisation with channel estimation, e.g. determination of delay spread, derivative or peak tracking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/309—Measuring or estimating channel quality parameters
    • H04B17/336—Signal-to-interference ratio [SIR] or carrier-to-interference ratio [CIR]
    • 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
    • H04L25/025—Channel estimation channel estimation algorithms using least-mean-square [LMS] method
    • 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
    • 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
    • H04L25/0256—Channel estimation using minimum mean square error criteria

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Power Engineering (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Electromagnetism (AREA)
  • Quality & Reliability (AREA)
  • Monitoring And Testing Of Transmission In General (AREA)
  • Noise Elimination (AREA)

Abstract

The invention discloses a channel interpolation method based on a neural network, which mainly solves the problem that the traditional linear interpolation and Lagrange interpolation method has performance bottleneck under high signal-to-noise ratio. The implementation scheme is as follows: 1) constructing a neural network; 2) collecting a training data set for a neural network; 3) performing offline training on the neural network by using a training data set; 4) after the neural network training is finished, obtaining the frequency domain estimation vector on the known frequency point of the receiving end again; 5) and inputting the frequency domain estimation vectors on the known frequency points into the trained neural network to obtain the frequency domain estimation vectors on all the frequency points, thereby finishing channel estimation. Because the neural network is introduced during channel estimation, compared with the traditional algorithm, the method can more accurately estimate the channel information on unknown frequency points, does not generate performance bottleneck under the condition of high signal-to-noise ratio, improves the channel estimation precision, and can be used for channel estimation of the orthogonal frequency division multiplexing OFDM technology.

Description

Channel interpolation method based on neural network
Technical Field
The invention belongs to the technical field of communication, and particularly relates to a channel interpolation method which can be used for channel estimation of Orthogonal Frequency Division Multiplexing (OFDM).
Background
With the arrival of the 5G era, application scenarios of wireless communication are more and more diversified, besides traditional communication services, smart homes, smart cities, automatic driving, cloud office and the like are provided, the diversified application scenarios bring higher performance index requirements, and the traditional communication technology is difficult to meet the increasing requirements in the aspects of transmission rate, transmission performance and the like.
The rapid development of artificial intelligence brings a new research direction to the field of communication technology, the basis of the artificial intelligence technology is a neural network technology, the neural network can effectively approach and fit any complex function, and can extract and process implicit characteristics, so that the neural network technology has great application potential in the physical layer of communication. The paper "Power of Deep Learning for Channel Estimation and Signal Detection in OFDM systems" published by Hao Ye et al in IEEE Wireless Communications Letters,2018,7(1): 114-.
In the present phase, many schemes for channel estimation using neural networks have been developed. The paper "Deep Learning-Based Channel Estimation and Equalization Scheme for FBMC/OQAM Systems" published by Xing Cheng et al in "IEEE Wireless Communications Letters,2019,8(3): 881-884" combines the Channel Estimation and Equalization modules and replaces them with a neural network, resulting in better performance than the conventional method, but the algorithm does not separate the Channel Estimation module separately, which is not beneficial to the modularization of the communication system.
The traditional channel interpolation method comprises a linear interpolation method and a Lagrange interpolation method, wherein the linear interpolation method is simple in calculation, but the interpolation performance is poor; under the condition of high-order interpolation, the performance of the Lagrange interpolation method is superior to that of a linear interpolation method, but the higher the interpolation times is, the higher the calculation complexity is. The "phase filter-based OFDM transmission system" published by Chang L S et al in "IEEE Vehicular Technology Conference,2004,1: 525-. In addition, the interpolation method has a bottleneck along with the improvement of the signal-to-noise ratio, and the performance gap with the ideal channel estimation is larger and larger.
Disclosure of Invention
The invention aims to provide a channel interpolation method based on a neural network to more accurately estimate channel information on all frequency points according to the channel information on known frequency points and improve the accuracy of channel estimation aiming at the defects of the prior art.
The technical idea of the invention is as follows: combining a neural network with a traditional least square channel estimation scheme, namely after the traditional least square channel estimation scheme, estimating channel information on all frequency points by using the neural network to obtain frequency domain channel information, wherein the implementation steps comprise the following steps:
1) building a neural network model comprising an input layer, three hidden layers and an output layer which are sequentially connected, wherein the two adjacent layers are connected in a full connection mode;
2) setting a random multipath channel, and recording a frequency domain channel vector H (n) of the channel, wherein n is 0,1,2,. and 255;
3) after the transmitted signal passes through the multipath channel, least square estimation LS is carried out at the receiving end to obtain the frequency domain estimation vector of the known frequency point
Figure BDA0002594410960000021
Wherein p is 0,1,2,. times, 31;
4) frequency domain estimation vector for known frequency points
Figure BDA0002594410960000022
And frequency domain channel vector H (n) of multipath channel, making data pretreatment of real part and imaginary part separation to form a training sample;
5) repetition of 2) to 4) of N in totaldThen, obtaining a mixture containing NdTraining data set of group training samples, where Nd＝400000；
6) Substituting the training data set into the built neural network to perform off-line training to obtain a trained neural network;
7) performing least square estimation at a receiving end to obtain a frequency domain channel estimation vector on a known frequency point
Figure BDA0002594410960000023
And after the pretreatment of separating real part from imaginary partInputting the frequency domain channel estimation vectors into the neural network trained in the step 6) to obtain frequency domain channel estimation vectors on all frequency points
Figure BDA0002594410960000024
Compared with the prior art, the invention has the following advantages:
1. on the basis of the traditional least square channel estimation scheme, the invention introduces a neural network aiming at different multipath channels, can more accurately estimate the channel information on all frequency points according to the channel information on the known frequency points, and can not generate the bottleneck of interpolation performance along with the increase of the signal-to-noise ratio, thereby improving the performance of channel estimation.
2. The invention independently applies the neural network to the channel estimation process, and is beneficial to the modular design of the communication system.
Drawings
FIG. 1 is a flow chart of an implementation of the present invention;
FIG. 2 is a diagram of a neural network constructed in accordance with the present invention;
fig. 3 is a graph comparing the bit error rate performance of the present invention with that of the conventional channel interpolation method.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific embodiments.
Referring to fig. 1, the implementation steps of this example are as follows:
step 1, constructing a neural network.
1.1) neural network architecture:
referring to fig. 2, the neural network model constructed in this step includes a 5-layer structure, which is sequentially: the input layer → the first hidden layer → the second hidden layer → the third hidden layer → the output layer, the two adjacent layers are connected by using a full connection mode, namely the output end of the input layer neuron is connected with the input end of the first hidden layer neuron in pairs, the output end of the first hidden layer neuron is connected with the input end of the second hidden layer neuron in pairs, the output end of the second hidden layer neuron is connected with the input end of the third hidden layer neuron in pairs, and the output end of the third hidden layer neuron is connected with the input end of the output layer neuron in pairs;
1.2) parameters of each layer:
the input layer comprises 64 neurons, and the output value of each neuron is sequentially represented as
Figure BDA0002594410960000031
Wherein i is 1,2,31，n1＝64；
The first hidden layer comprises 100 neurons, and the output value of each neuron is sequentially expressed as
Figure BDA0002594410960000032
The calculation formula is as follows:
Figure BDA0002594410960000033
wherein j is 1,2,32，n2＝100；
Figure BDA0002594410960000034
A weight parameter for the first hidden layer;
Figure BDA0002594410960000035
a bias parameter for a first hidden layer; sigma2() is an activation function and the first hidden layer uses a sigmoid function, the formula is:
Figure BDA0002594410960000036
the second hidden layer comprises 250 neurons, and the output value of each neuron is sequentially expressed as
Figure BDA0002594410960000037
The calculation formula is as follows:
Figure BDA0002594410960000038
wherein k is 1,2,33，n3＝250；
Figure BDA0002594410960000039
A weight parameter for the second hidden layer;
Figure BDA00025944109600000310
a bias parameter for the second hidden layer; sigma3(. cndot.) is an activation function, and the second hidden layer uses a sigmoid function, and the formula is
Figure BDA0002594410960000041
The third hidden layer comprises 500 neurons, and the output value of each neuron is sequentially represented as
Figure BDA0002594410960000042
The calculation formula is as follows:
Figure BDA0002594410960000043
wherein, l is 1,2,34，n4＝500；
Figure BDA0002594410960000044
A weight parameter of a third hidden layer;
Figure BDA0002594410960000045
a bias parameter for a third hidden layer; sigma4() is an activation function, and the third hidden layer uses a sigmoid function, and the formula is:
Figure BDA0002594410960000046
the output layer comprises 512 neurons, and the output value of each neuron is sequentially expressed as
Figure BDA0002594410960000047
The calculation formula is as follows:
Figure BDA0002594410960000048
wherein, m is 1,2,35，n5＝512；
Figure BDA0002594410960000049
Is the weight parameter of the output layer;
Figure BDA00025944109600000410
is the bias parameter of the output layer; sigma5(. cndot.) is the activation function, the output layer uses a linear function, expressed as: sigma5(x)＝x；
1.3) two parameter sets of the neural network:
using weight parameters in neural networks
Figure BDA00025944109600000411
Forming a weight parameter set W;
using bias parameters in neural networks
Figure BDA00025944109600000412
Constituting a bias parameter set B.
And 2, acquiring a training data set.
2.1) setting a multipath channel model:
firstly, randomly setting the path number of a multipath channel model to be 2-10 paths, wherein the attenuation range of each path is random 0-20 dB; then randomly generating 0-0.62 microsecond multipath time delay, and recording the frequency domain channel vector H (n) of the channel;
then, selecting a signal-to-noise ratio to represent the amount of Gaussian white noise added into the multipath channel;
the signal-to-noise ratio selected in this example is 20 dB;
2.2) after the transmitted signal has passed through the multipath channelThe receiving end firstly carries out least square estimation LS to obtain a frequency domain channel estimation vector on a known frequency point
Figure BDA00025944109600000413
The formula is as follows:
Figure BDA0002594410960000051
wherein x (p) is a known transmit pilot, and y (p) is a receive pilot;
2.3) estimating vector for time domain channel on known frequency point
Figure BDA0002594410960000052
Performing a pretreatment, i.e. separation
Figure BDA0002594410960000053
The real part and the imaginary part of each element in the time domain channel estimation vector are obtained
Figure BDA0002594410960000054
And imaginary part of time domain channel estimation vector
Figure BDA0002594410960000055
Then will be
Figure BDA0002594410960000056
And
Figure BDA0002594410960000057
performing serial connection to obtain a real number estimation vector
Figure BDA0002594410960000058
2.4) preprocessing the frequency domain channel vector H (n) of the channel in 2.1), i.e. separating the real part and the imaginary part of each element in H (n) to obtain the real part H of the frequency domain channel vectorR(n) and imaginary part H of channel vector in frequency domainI(n) adding HR(n) and HI(n) are concatenated to obtain a real number channel vectorH′(n)；
2.5) estimating the vector with real numbers
Figure BDA0002594410960000059
The real channel vector H' (n) is a label and forms a training sample;
2.6) repetition of 2.1) to 2.5) with NdThen, obtaining a mixture containing NdTraining data set of group training samples, N in this exampled＝400000。
And 3, substituting the data set into a neural network for off-line training.
3.1) will contain NdThe training data set of the set of training samples is separated into two parts, a training set and a test set, where NtrGroup as training set, N remainsteGroup as test set, Nd＝Ntr+NteIn this example Ntr＝300000，Nte＝100000；
3.2) selecting a loss function J, wherein the loss function selected in the example is a mean square error function, and the formula is as follows:
Figure BDA00025944109600000510
wherein, ymIs the output of the neural network and is,
Figure BDA00025944109600000511
labels that are training samples;
3.3) training the neural network by adopting a random gradient descent method:
3.3.1) initializing neural networks: from [0,1 ]]Selecting a weight parameter set W and a bias parameter set B from uniformly distributed random numbers, and setting a threshold value J of a loss function0；
3.3.2) randomly selecting 10 samples in the training set, and respectively substituting the samples into an input layer of the neural network to carry out operation to obtain 10 output values;
3.3.3) substituting the 10 output values in 3.3.2) and the labels in the training set into a loss function respectively, calculating a loss function value, and calculating the loss function valueAveraging the function values to obtain an average value
Figure BDA00025944109600000512
3.3.4) average value
Figure BDA00025944109600000513
And threshold value J0Make a comparison if
Figure BDA00025944109600000514
Finishing the training to obtain a trained neural network, and otherwise, executing 3.3.5);
3.3.5) performing back propagation operation, namely respectively solving partial derivatives of all parameters in the weight parameter set W and the bias parameter set B by using a loss function J to obtain a weight gradient set delta W and a bias gradient set delta B, wherein elements in the weight gradient set delta W correspond to elements in the weight parameter set W one by one, and elements in the bias gradient set delta B correspond to elements in the bias parameter set B one by one;
3.3.6) updating the set of weight parameters W and the set of bias parameters B according to the set of weight gradients Δ W and the set of bias gradients Δ B as:
Figure BDA00025944109600000613
w' is the updated weight parameter set; b' is the updated bias parameter set; η is a learning rate, where η (0) is set to 0.1 as an initial value, and the subsequent value is attenuated according to the number of rounds of training, and the attenuation formula is:
Figure BDA0002594410960000061
in the formula, q is the number of times of updating the parameters, d is an attenuation factor, and the value is 0.0001;
3.3.7) repeating 3.3.2) to 3.3.6) until the data in the training set is completely extracted, completing a training round;
3.3.8) repeating the steps from 3.3.2) to 3.3.7) until the loss function value meets the requirement of 3.3.4), stopping training and obtaining the trained neural network.
And 4, performing interpolation at the receiving end by using the trained neural network.
4.1) obtaining the frequency domain receiving signal at the receiving end to carry out least square channel estimation to obtain the frequency domain estimation vector of the known frequency point
Figure BDA0002594410960000062
Figure BDA0002594410960000063
Wherein x '(p) is a known transmitting pilot frequency after the neural network is trained, and y' (p) is a receiving pilot frequency after the neural network is trained;
4.2) estimating vector for time domain channel on known frequency point
Figure BDA0002594410960000064
Performing a pretreatment, i.e. separation
Figure BDA0002594410960000065
The real part and the imaginary part of each element in the time domain channel estimation vector are obtained
Figure BDA0002594410960000066
And imaginary part of time domain channel estimation vector
Figure BDA0002594410960000067
Will be provided with
Figure BDA0002594410960000068
And
Figure BDA0002594410960000069
performing serial connection to obtain a real number estimation vector
Figure BDA00025944109600000610
4.3) estimating the vector of real numbers
Figure BDA00025944109600000611
Inputting the frequency domain estimation vector on all frequency points into a trained neural network, wherein the output of the neural network is the frequency domain estimation vector on all frequency points
Figure BDA00025944109600000612
At this point, channel estimation is completed.
The effects of the present invention can be further illustrated by the following simulations:
1. simulation conditions
The transmission system used for simulation is an Orthogonal Frequency Division Multiplexing (OFDM) system; the sampling rate of the system is 50MHz, the number of subcarriers is 256, and the length of a cyclic prefix is 64; the channel is an extended pedestrian channel model EPA, the channel comprises 7 paths, the power attenuation is respectively 0.0, -1.0, -2.0, -3.0, -8.0, -17.2 and-20.8 dB, and the multipath time delay is respectively 0, 30, 70, 90, 110, 190 and 410 nanoseconds; the pilot frequency in the data frame adopts a scattered pilot frequency structure. The standard for measuring the simulation result is the bit error rate, i.e. the ratio of the number of bits of system transmission errors to the total number of bits of transmission.
2. Emulated content
The bit error rate simulation is performed by using the method of the invention and two traditional channel interpolation methods respectively, and the result is shown in figure 3.
The abscissa of fig. 3 is the signal-to-noise ratio and the ordinate is the bit error rate of the system. Wherein:
the IDEAL curve refers to the performance of an IDEAL channel estimate, which represents the ultimate performance of the system when the channel estimate is completely error free;
the LS-linear interpolation curve is a bit error rate curve using the existing linear interpolation algorithm;
the LS-Lagrange quadratic interpolation curve is the error rate of the existing Lagrange quadratic interpolation algorithm;
the NN-LS curve is the error rate curve of the invention;
comparing the error rate performance of the invention and the traditional channel interpolation algorithm, it can be found that the invention has better interpolation effect and better system performance than the traditional linear interpolation and Lagrange quadratic interpolation method when the signal-to-noise ratio is more than 15 dB.

Claims (6)

1. A channel interpolation method based on a neural network is characterized by comprising the following steps:
1) building a neural network model which comprises an input layer, three hidden layers and an output layer which are sequentially connected, wherein the two adjacent layers are connected in a full connection mode;
2) setting a random multipath channel, and recording a frequency domain channel vector H (n) of the channel, wherein n is 0,1,2,. and 255;
3) after the transmitted signal passes through the multipath channel, least square estimation LS is carried out at the receiving end to obtain the frequency domain estimation vector of the known frequency point
Figure FDA0003159695460000011
Wherein p is 0,1,2,. times, 31;
4) frequency domain estimation vector for known frequency points
Figure FDA0003159695460000012
And frequency domain channel vector H (n) of multipath channel, making data pretreatment of real part and imaginary part separation to form a training sample; the method is realized as follows:
4a) estimating vectors for frequency domain
Figure FDA0003159695460000013
The real part and the imaginary part of each element are separated to obtain the real part of the frequency domain estimation vector
Figure FDA0003159695460000014
Imaginary part of sum frequency domain estimation vector
Figure FDA0003159695460000015
Will be provided with
Figure FDA0003159695460000016
And
Figure FDA0003159695460000017
concatenated into a real estimate vector
Figure FDA0003159695460000018
4b) Separating the real part and the imaginary part of each element in the frequency domain channel vector H (n) to obtain the real part H of the frequency domain channel vectorR(n) and imaginary part H of channel vector in frequency domainI(n) reacting HR(n) and HI(n) concatenating into a real channel vector H' (n);
4c) estimating vectors by real numbers
Figure FDA0003159695460000019
The real channel vector H' (n) is a label and forms a training sample;
5) repetition of 2) to 4) of N in totaldThen, obtaining a mixture containing NdTraining data set of group training samples, where Nd＝400000；
6) Substituting the training data set into the built neural network to perform off-line training to obtain a trained neural network;
7) performing least square estimation at a receiving end to obtain a frequency domain channel estimation vector on a known frequency point
Figure FDA00031596954600000110
Preprocessing the real part and the imaginary part of the signal, inputting the preprocessed signal into the neural network trained in the 6) to obtain frequency domain channel estimation vectors on all frequency points
Figure FDA00031596954600000111
2. The method of claim 1, wherein the neural network model constructed in 1) has the following parameters for each layer:
an input layer comprising 64 neurons;
the first hidden layer comprises 100 neurons, and the used activation function is a sigmoid function;
a second hidden layer, which comprises 250 neurons and uses a sigmoid function as an activation function;
a third hidden layer which comprises 500 neurons and uses an activation function which is a sigmoid function;
the output layer, which contains 512 neurons, uses an activation function that is a linear function.
3. The method according to claim 1, wherein the 2) is configured with a random multipath channel, the number of paths of the multipath channel model is randomly set to 2 to 10 paths, and the attenuation range of each path is randomly 0-20 dB; then randomly generating 0-0.62 microsecond multipath time delay, and recording the frequency domain channel vector H (n) of the channel; then, a signal-to-noise ratio is selected, which represents the amount of white gaussian noise added by the multipath channel.
4. The method of claim 1, wherein the frequency domain estimation vectors at known frequency points obtained in 3)
Figure FDA0003159695460000021
Is represented as follows:
Figure FDA0003159695460000022
where x (p) is the known transmit pilot and y (p) is the pilot portion of the received signal.
5. The method of claim 1, wherein the neural network is trained offline in 6) by:
6a) will contain NdThe training data set of the set of training samples is separated into two parts, where NtrGroup as training set, N remainsteGroup as test set, Nd＝Ntr+Nte；
6b) Selecting a mean square error function as a loss function J;
6c) training the neural network by adopting a random gradient descent method:
6c1) initializing a neural network: from [0,1 ]]Selecting a weight parameter set W and a bias parameter set B from uniformly distributed random numbers, and setting a threshold value J of a loss function0；
6c2) Randomly selecting 10 samples in a training set, and respectively substituting the samples into an input layer of a neural network to carry out operation to obtain 10 output values;
6c3) respectively substituting 10 output values in 6c2) and labels in the training set into a loss function to calculate loss function values, and averaging the loss function values to obtain an average value
Figure FDA0003159695460000023
6c4) Average value
Figure FDA0003159695460000024
And a threshold value J0Make a comparison if
Figure FDA0003159695460000025
Training is complete, otherwise, 6c5 is executed);
6c5) performing back propagation operation, namely solving partial derivatives of all parameters in the weight parameter set W and the bias parameter set B by using a loss function J to obtain a weight gradient set delta W and a bias gradient set delta B, wherein elements in the weight gradient set delta W correspond to elements in the weight parameter set W one by one, and elements in the bias gradient set delta B correspond to elements in the bias parameter set B one by one;
6c6) according to the weight gradient set Δ W and the bias gradient set Δ B, the weight parameter set W and the bias parameter set B are updated as follows:
Figure FDA0003159695460000031
w' is the updated weight parameter set; b' is the updated bias parameter set; η is a learning rate, and η (0) is an initial value of 0.1, and then as the number of rounds of training is attenuated, the attenuation formula is:
Figure FDA0003159695460000032
wherein d is an attenuation factor, 0.0001 is taken, and q represents the number of times that the parameter updating is carried out;
6c7) repeating 6c2) to 6c6) until the data in the training set is completely taken, and completing a round of training;
6c8) repeat 6c2) to 6c7) until the loss function value meets the requirement in 6c4), stop training.
6. The method of claim 1, wherein the frequency domain estimation vectors at known frequency points obtained in 7)
Figure FDA0003159695460000033
Is represented as follows:
Figure FDA0003159695460000034
wherein x '(p) is the known transmitted pilot after the neural network has been trained, and y' (p) is the received pilot after the neural network has been trained.
CN202010705053.9A 2020-07-21 2020-07-21 Channel interpolation method based on neural network Active CN111884976B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010705053.9A CN111884976B (en) 2020-07-21 2020-07-21 Channel interpolation method based on neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010705053.9A CN111884976B (en) 2020-07-21 2020-07-21 Channel interpolation method based on neural network

Publications (2)

Publication Number Publication Date
CN111884976A CN111884976A (en) 2020-11-03
CN111884976B true CN111884976B (en) 2021-09-03

Family

ID=73155680

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010705053.9A Active CN111884976B (en) 2020-07-21 2020-07-21 Channel interpolation method based on neural network

Country Status (1)

Country Link
CN (1) CN111884976B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023104317A1 (en) * 2021-12-10 2023-06-15 Nokia Solutions And Networks Oy A radio receiver device with a neural network, and related methods and computer programs

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112737987B (en) * 2020-12-28 2022-11-01 南京邮电大学 Novel time-varying channel prediction method based on deep learning
CN112906035B (en) * 2021-03-24 2022-11-18 东南大学 Method for generating frequency division duplex system key based on deep learning
CN113572708B (en) * 2021-06-30 2023-03-14 西安电子科技大学 DFT channel estimation improvement method
CN115664898B (en) * 2022-10-24 2023-09-08 四川农业大学 OFDM system channel estimation method and system based on complex convolution neural network
CN117057407A (en) * 2023-08-21 2023-11-14 浙江大学 Training method for crosstalk-oriented wavelength division multiplexing optical neural network

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103414668A (en) * 2013-08-30 2013-11-27 西安电子科技大学 Method for estimating channel coefficient of two-way relay cooperative system based on training sequence
CN109194595A (en) * 2018-09-26 2019-01-11 东南大学 A kind of adaptive OFDM method of reseptance of channel circumstance neural network based

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107135041B (en) * 2017-03-28 2020-12-29 西安电子科技大学 RBF neural network channel prediction method based on phase space reconstruction
US10601471B1 (en) * 2018-08-29 2020-03-24 Micron Technology, Inc. Neuron calculator for artificial neural networks
CN109617847B (en) * 2018-11-26 2021-04-06 东南大学 OFDM receiving method without cyclic prefix based on model-driven deep learning
CN109743268B (en) * 2018-12-06 2022-02-15 东南大学 Millimeter wave channel estimation and compression method based on deep neural network
CN109672464B (en) * 2018-12-13 2021-09-03 西安电子科技大学 FCFNN-based large-scale MIMO channel state information feedback method
CN111404849B (en) * 2020-03-20 2021-01-12 北京航空航天大学 OFDM channel estimation and signal detection method based on deep learning

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103414668A (en) * 2013-08-30 2013-11-27 西安电子科技大学 Method for estimating channel coefficient of two-way relay cooperative system based on training sequence
CN109194595A (en) * 2018-09-26 2019-01-11 东南大学 A kind of adaptive OFDM method of reseptance of channel circumstance neural network based

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023104317A1 (en) * 2021-12-10 2023-06-15 Nokia Solutions And Networks Oy A radio receiver device with a neural network, and related methods and computer programs

Also Published As

Publication number Publication date
CN111884976A (en) 2020-11-03

Similar Documents

Publication Publication Date Title
CN111884976B (en) Channel interpolation method based on neural network
CN111614584B (en) Transform domain adaptive filtering channel estimation method based on neural network
CN111683024B (en) Time-varying OFDM system channel estimation method based on deep learning
CN108566257B (en) Signal recovery method based on back propagation neural network
CN112737987B (en) Novel time-varying channel prediction method based on deep learning
CN112600772A (en) OFDM channel estimation and signal detection method based on data-driven neural network
CN113381828B (en) Sparse code multiple access random channel modeling method based on condition generation countermeasure network
CN113472706A (en) MIMO-OFDM system channel estimation method based on deep neural network
CN112115821B (en) Multi-signal intelligent modulation mode identification method based on wavelet approximate coefficient entropy
CN114268388B (en) Channel estimation method based on improved GAN network in large-scale MIMO
CN113572708B (en) DFT channel estimation improvement method
CN113285896B (en) Time-varying channel prediction method based on stack type ELM
CN110311876A (en) The implementation method of underwater sound OFDM receiver based on deep neural network
CN110138698A (en) High order modulation linear hybrid signal frequency deviation first phase combined estimation method and device
CN112422208B (en) Signal detection method based on antagonistic learning under unknown channel model
CN111131108A (en) Non-cooperative underwater acoustic OFDM subcarrier modulation mode identification method
CN110944002B (en) Physical layer authentication method based on exponential average data enhancement
Li et al. A novel channel estimation method based on deep neural network for otfs system
CN114614920B (en) Signal detection method based on data and model combined driving of learning factor graph
CN113489545B (en) Light space pulse position modulation step-by-step classification detection method based on K-means clustering
CN110233808A (en) A kind of FTN system signal detection method
CN112821971A (en) Time-varying channel signal detection method based on countermeasure learning
CN113890633A (en) Underwater acoustic communication system self-adaptive selection method based on deep neural network
CN117278363A (en) OFDM system frequency offset estimation algorithm based on deep learning in fast time-varying scene
Fathi et al. Deep Learning For Channel Estimation And Signal Detection

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