CN109067688A - A kind of OFDM method of reseptance of data model double drive - Google Patents
A kind of OFDM method of reseptance of data model double drive Download PDFInfo
- Publication number
- CN109067688A CN109067688A CN201810743534.1A CN201810743534A CN109067688A CN 109067688 A CN109067688 A CN 109067688A CN 201810743534 A CN201810743534 A CN 201810743534A CN 109067688 A CN109067688 A CN 109067688A
- Authority
- CN
- China
- Prior art keywords
- channel estimation
- neural network
- deep neural
- reseptance
- data model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
-
- 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
- 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
- H04L27/2692—Link with other circuits, i.e. special connections between synchronisation arrangements and other circuits for achieving synchronisation with preamble design, i.e. with negotiation of the synchronisation sequence with transmitter or sequence linked to the algorithm used at the receiver
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Power Engineering (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
The invention discloses a kind of OFDM method of reseptances of data model double drive, comprising: is trained respectively to the deep neural network in channel estimation and signal detection module in receiver；For channel estimation module, input received pilot tone and local pilot tone, estimate to obtain LS channel estimation using the least square estimation method and improves as a result, real and imaginary parts is taken to take out series connection input deep neural network and export channel estimation results；For signal detection module, input the channel estimation results that received pilot tone and channel estimation module obtain, zero forcing equalization is obtained using zero forcing equalization method as a result, and input deep neural network improve and exported；Gained deep neural network is exported and does hard decision with setting thresholding and obtains court verdict, and obtains sending the estimation of bit stream by court verdict.The present invention is optimized and is improved by neural network, and the time loss of network training is few, and detection performance is high, can be quickly obtained channel information.
Description
Technical field
The present invention relates to a kind of OFDM method of reseptance methods of data model double drive, belong to wireless communication technology field.
Background technique
In recent years, deep learning is as the basic technology in artificial intelligence, in computer vision and natural language processing etc.
Subject achieves immense success, surmounts under the scenes such as image classification, face recognition, speech recognition, machine translation, style conversion
The performance of conventional machines learning methods, so that unmanned, intelligent medical diagnosis on disease, smart home, personalized recommendation etc. are applied
Become possibility.Deep learning is a branch in machine learning field, is a kind of method of supervised learning, deep by minimizing
The loss function between the predicted value and true value of neural network is spent, one group of optimal neural network parameter is obtained, is come so that deep
Degree neural network is able to carry out Accurate Prediction.
The research for having had some explorations that deep learning is applied in wireless communication physical layer, including entire communication system
System is substituted by deep neural network end to end completely, or the part of module of communication system is only replaced by deep neural network,
Such as encoder, decoder, detector etc..But the knowledge of wireless communication field is not applied to by currently used method
In neural network design, so that the function of neural network is entirely a black box, place one's entire reliance upon greatly to the training of neural network
Data-driven is measured, the parameter amount of neural network is also bigger, and training speed is slow.
Summary of the invention
Technical problem to be solved by the present invention lies in overcome the deficiencies of the prior art and provide a kind of data model double drive
OFDM method of reseptance, solve of the existing technology communication not being added when the communication control processor in deep learning network application
Knowledge and the problem of be completely dependent on data-driven.
The present invention specifically uses following technical scheme to solve above-mentioned technical problem:
A kind of OFDM method of reseptance of data model double drive, comprising the following steps:
Step 1 respectively instructs the deep neural network in channel estimation module and signal detection module in receiver
Practice；
Step 2, for channel estimation module, received pilot tone and local pilot tone are inputted, using least square
Estimation method estimation obtains LS channel estimation resultAnd LS channel estimation knot real and imaginary parts is taken to take out
The deep neural network trained is inputted after series connection to improve and export to obtain channel estimation results
Step 3, for signal detection module, input the channel estimation that received pilot tone and channel estimation module obtain
As a resultZero forcing equalization result is obtained using zero forcing equalization methodAnd by zero forcing equalization resultReal and imaginary parts take
Series connection and channel estimation results outThe deep neural network that signal input has been trained is received to improve and exported；
Step 4 exports step 3 gained deep neural network and does hard decision with setting thresholding and obtain court verdict, and
Obtain sending the estimation of bit stream by court verdict.
Further, as a preferred technical solution of the present invention, deep neural network training is adopted in the step 1
Loss function is square error loss, and the optimizer used is adaptability momentum Estimation Optimization device.
Further, as a preferred technical solution of the present invention, the loss function used in the step 1 is square
Error loss, specifically includes:
In channel estimation module, the square error loss of use is:
In signal detection module, the square error loss of use is:
Wherein, N is sub-carrier number, and B is the bit number to be estimated, and H is actual frequency domain channel,It is channel estimation knot
Fruit, b are practical transmission bit streams,It is the estimation for sending bit stream.
Further, as a preferred technical solution of the present invention, LS channel estimation is obtained in the step 2
As a resultUsing formula:
Wherein, Y ' is received pilot tone, and X ' is local pilot tone.
The present invention by adopting the above technical scheme, can have the following technical effects:
Communicating knowledge is reasonably dissolved into neural network design by the present invention, and performance is equal under linear and nonlinear situation
It is promoted obviously, net training time greatly shortens, and suitable for the system using OFDM, the present invention instructs OFDM receiver sub-module
Practice neural network, and uses the result of traditional communication algorithm to be input to neural network as initial value and optimize and improve, network
Trained time loss is few, and detection performance is high.In addition, when environment is more complicated, it is available steady by series connection conventional module
Fixed performance.
Therefore, the present invention is provided by being combined deep learning and communicating knowledge for the design of ofdm system receiver
A kind of new method, in the case of linear and non-linear the method is than conventional linear method of reseptance and full connection method BER
Can have a promotion, and it is fewer than full connection method parameter, training speed is fast, can get channel information, training, line top under this line
The scheme of administration is also suitable for realizing and applies.
Detailed description of the invention
Fig. 1 is the OFDM receiver block diagram of data model double drive of the invention.
Fig. 2 is channel estimation module block diagram of the invention.
Fig. 3 is deep neural network block diagram in the signal detection module under nonlinear situation of the invention.
Specific embodiment
Embodiments of the present invention are described with reference to the accompanying drawings of the specification.
As shown in Figure 1, being the schematic diagram of the OFDM receiver of data model double drive, the present invention proposes that one kind can be used for this
The OFDM method of reseptance of the data model double drive of receiver, but be not limited to this kind of receiver structure, this method specifically include with
Lower step:
Step 1 respectively instructs the deep neural network in channel estimation module and signal detection module in receiver
Practice.The deep neural network of data model double drive is applied in wireless communication receiver, online lower supervised learning
Method is trained network, carries out forward prediction with trained network in practical application, and the specific structure of receiver is
It is divided into two submodules of channel estimation and signal detection, first the deep neural network in channel estimation module is instructed when training
White silk is again trained the deep neural network in signal detection module, and when application first obtains the channel of channel estimation module output
Estimated result is entered into signal detection module again.
Wherein, the loss function used to the training of deep neural network is square error loss, and the optimizer of use is
Adaptability momentum estimates Adam optimizer, is estimated using the learning rate of dynamic adjustment in channel using the training method of small lot
It counts in module, square error loss is:
In signal detection module, square error loss is:
Wherein, N is sub-carrier number, and B is the bit number to be estimated, and H is actual frequency domain channel,It is channel estimation knot
Fruit, b are practical transmission bit streams,It is the estimation for sending bit stream.
Step 2, for channel estimation module, received pilot tone and local pilot tone are inputted, using least square
Estimation method is initialized, and for k-th of subcarrier, is obtained and is estimated to obtain LS channel estimation resultAre as follows:
Wherein, Y ' is received pilot tone, and X ' is local pilot tone.
Then by LS channel estimation resultReal and imaginary parts take out series connection, be sent into connect entirely depth mind
It is improved through network, exports channel estimation results
In this step, the fully-connected network that deep neural network is one layer does not use activation primitive, the multiplying property of the layer network
Parameter W is initialized using the real value of the weight matrix of linear minimum mean-squared error channel estimation, and additivity parameter n is initialized as
Complete zero, and set trainable for all parameters, network implementations is calculated as at this timeBecause of lowest mean square
Error signal estimationAnd real-valued calculation can only be carried out in deep learning network, so working as the network
Input beReal and imaginary parts series connection, the multiplying property parameter W of network is initialized as weight matrix WLMMSEReal value matrixWherein, Re { } is to take real part, and Im { } is to take imaginary part；Additivity parameter
When n is initialized as complete zero, the initial value of network output isReal and imaginary parts series connection, then by network parameter into
Row training, the channel estimation results that network can be made to exportAccuracy is more than
Step 3, for signal detection module, input the channel estimation that received pilot tone and channel estimation module obtain
As a resultIt is initialized using zero forcing equalization method, for k-th of subcarrier, obtains zero forcing equalization resultAre as follows:
Wherein, Y is received frequency domain data,It is the output of channel estimation module.
And by zero forcing equalization resultReal and imaginary parts take out series connection and channel estimation resultsReceive signal input
The deep neural network trained is improved and is exported.Wherein, receiving signal can be the user data received.
Wherein, in signal detection module, can based on system be it is linear or non-linear, use is two different
Deep neural network:
When ofdm system is linear system, using the deep neural network connected entirely；The input of the network is that force zero is equal
Weighing apparatus as a result, network uses two layers of fully-connected network.
When ofdm system is nonlinear system, such as remove cyclic prefix to which there are the systems of intersymbol interference, or
Person is to reduce the equal energy ratio in peak to do truncation there are the signal of truncation noise, and signal detection improvement network uses two-way long short-term memory
BiLSTM Recognition with Recurrent Neural Network and one layer of full Connection Neural Network.The input of the network is estimated for the result of zero forcing equalization, channel
It counts the channel estimation results of module output, receive signal, network is using three layers, the BiLSTM of OFDM number of carrier wave time phase
Then the output of BiLSTM Recognition with Recurrent Neural Network is sent into one layer of fully-connected network by Recognition with Recurrent Neural Network.
And the wherein mind of signal detection module deep neural network the last layer when ofdm system is nonlinear system
It is equal to the bit number B to be estimated through first number, is determined by the part amount of bits that continuously subcarrier is included, meeting
Change with the constellation order of modulation difference of use.
Step 4 exports step 3 gained deep neural network and does hard decision with setting thresholding and obtain court verdict, and
Obtain sending the estimation of bit stream by court verdict.
The present embodiment can output to deep neural network in signal detection module with 0.5 do hard decision for thresholding, be greater than
0.5 is judged to bit 1, less than 0.5 is judged to bit 0, obtains the estimation for sending bit stream.
When channel status is deteriorated or changes too fast, using traditional communication means, i.e., it will receive pilot tone and locally lead
Frequency meter calculates LS channel estimation, is then sent to the result of LS channel estimation and reception frequency domain data together urgent
Then zero detection module carries out constellation demapping to the result of squeeze theorem, obtains the general but stable transmission bit of accuracy
Stream estimation
Based on the above method, the present invention provides a specific embodiment.
As shown in Figure 1, by the Application of Neural Network of data model double drive in wireless communication receiver, instead of original
Channel estimation, signal equalization and the QAM of ofdm system demodulate the function of three modules, are specifically divided into channel estimation and signal inspection
Two submodules are surveyed, network are trained with the method for supervised learning under online, in practical application with trained network
Forward prediction is carried out, the specific structure of receiver is divided into two submodules of channel estimation and signal detection, first to letter when training
Deep neural network in road estimation module, which is trained, is again trained the deep neural network in signal detection module, answers
The channel estimation results that used time first obtains channel estimation module output are entered into signal detection module again.
As shown in Fig. 2, channel estimation module is inputted received pilot tone and local pilot tone, is first used
Least-squares estimation result is initialized, for k-th of subcarrier (1≤k≤64), LS channel estimationAre as follows:
Wherein, Y ' is received pilot tone, and X ' is local pilot tone, then by the real part of least-squares estimation result
The vector for being connected into that length is 128 is taken out with imaginary part, the deep neural network connected entirely is sent into and improves, and output length is
128 channel estimation resultsThe fully-connected network that deep neural network is one layer, does not use activation primitive, the layer network
Multiplying property parameter is initialized using the real value of weight matrix of linear minimum mean-squared error channel estimation, additivity parameter it is initial
Value is set as complete zero, and sets trainable for all parameters.Wherein least mean-square error channel estimationWeight matrix WLMMSEReal value matrixAre as follows:
Dimension is 128 × 128.To deep neural network training
Process be, setting label value be true frequency domain channel H, loss function be square error lossOptimizer is Adam optimizer, is declined when training using small lot gradient, is used in every wheel
50 batches, the size of each batch are 1000 samples, train 2000 wheels altogether, preceding using the learning rate of dynamic adjustment
1000 wheel learning rates are 0.001, and rear 1000 wheel learning rate is 0.0001.
For signal detection module, the channel estimation results of received frequency domain data and channel estimation module are inputted, are first adopted
It is initialized with zero forcing equalization result, for k-th of subcarrier, the result of zero forcing equalizationAre as follows:
Wherein, Y is received frequency domain data,It is the output of channel estimation module, then by the reality of zero forcing equalization result
Portion and imaginary part take out series connection or cochannel estimated resultReception signal is sent into deep neural network together and is improved, to letter
The output of deep neural network does hard decision with 0.5 thresholding in number detection module, bit 1 is judged to greater than 0.5, less than 0.5
Be judged to bit 0, output sends the estimation of bit stream
Based on ofdm system be it is linear or non-linear, using the depth nerve net of two different signal detection modules
Network.For linear system, deep neural network is using the deep neural network connected entirely；For being there are nonlinear effect
System, such as remove cyclic prefix to which there are the systems of intersymbol interference, or do to be truncated to exist for the reduction equal energy ratio in peak and cut
The signal of disconnected noise, deep neural network use BiLSTM Recognition with Recurrent Neural Network and one layer of full Connection Neural Network.
For the fully-connected network under linear system, inputs and beReal and imaginary parts series connection after length be 128 to
Amount exports the bit stream for being 48 for length and restores, and the neuronal quantity of two-tier network is respectively 120 and 48, the activation letter of use
Number is respectively ReLU and Sigmoid, and multiplying property parameter is initialized using the method for He initialization, and additivity parameter all initializes
It is zero.
For the BiLSTM Recognition with Recurrent Neural Network under nonlinear system, as shown in figure 3, using three layers of 64 time phase altogether
BiLSTM Recognition with Recurrent Neural Network, every layer of status number is respectively 20,10,6, and the output of the last layer BiLSTM is sent to
The fully-connected network that the activation primitive that one layer of neuron number is 48 is Sigmoid.K-th of time phase of first layer BiLSTM
Input are as follows:
Wherein Re { } is to take real part, and Im { } is to take imaginary part.
The process being trained to the signal detection improvement network under the above linear or nonlinear system is that label is arranged
Value sends bit stream b to be practical, and loss function is square error lossOptimizer is Adam excellent
Change device, declined when training using small lot gradient, 50 batches are used in every wheel, the size of each batch is 1000 samples.
5000 wheels are trained altogether, and using the learning rate of dynamic adjustment, initial learning rate is set as 0.001, and every 2000 wheel is reduced to
Current 1/5th.
When channel status is deteriorated or changes too fast, using traditional communication means, i.e., it will receive pilot tone and locally lead
Frequency meter calculates LS channel estimation, is then sent to the result of LS channel estimation and reception frequency domain data together urgent
Then zero detection module carries out constellation demapping to the result of squeeze theorem, obtains the general but stable transmission bit of accuracy
Stream estimation
Illustrated with the design parameter of experimental example, in the ofdm system containing 64 subcarriers, data frame format is one
Pilot OFDM symbols and an OFDM data symbol, pilot tone and data all take 64 subcarriers.The constellation modulation system of pilot tone
For QPSK, the constellation modulation system of data uses the 64QAM of LTE standard.In transmitting terminal, data bit has 64 × 6=384 ratio
Spy is converted into time domain and sends signal by the modulation of 64QAM constellation plus pilot tone framing, IFFT；After multipath channel, receiving
End does FFT transform and obtains frequency domain receives pilot tone and data, is sent into and of the invention is combined based on communicating knowledge and deep learning
Ofdm system receiver obtains the recovery of 8 × 6=48 bit in 8 continuous subcarriers, such as the 1 to 8th subcarrier or the 9th
To 16 subcarriers.The deep neural network of 64/8=8 independent signal detection modules of training, the output to deep neural network
Do hard decision, the recovery of available all 384 bits.
Therefore the present invention provides one by being combined deep learning and communicating knowledge for the design of ofdm system receiver
Kind new method, the method is more equal than conventional linear method of reseptance and full connection method BER performance in the case of linear and non-linear
Have a promotion, and it is fewer than full connection method parameter, training speed is fast, can get channel information, training under this line is disposed on line
Scheme is also suitable for realizing and applies.
Embodiments of the present invention are explained in detail above in conjunction with attached drawing, but the present invention is not limited to above-mentioned implementations
Mode within the knowledge of a person skilled in the art can also be without departing from the purpose of the present invention
It makes a variety of changes.
Claims (8)
1. a kind of OFDM method of reseptance of data model double drive, which comprises the following steps:
Step 1 is respectively trained the deep neural network in channel estimation module and signal detection module in receiver；
Step 2, for channel estimation module, received pilot tone and local pilot tone are inputted, using least-squares estimation
Method estimation obtains LS channel estimation resultAnd LS channel estimation knot real and imaginary parts is taken to take out series connection
The deep neural network trained is inputted afterwards to improve and export to obtain channel estimation results
Step 3, for signal detection module, input the channel estimation results that received pilot tone and channel estimation module obtainZero forcing equalization result is obtained using zero forcing equalization methodAnd by zero forcing equalization resultReal and imaginary parts take out string
Connection and channel estimation resultsThe deep neural network that signal input has been trained is received to improve and exported；
Step 4 exports step 3 gained deep neural network and does hard decision with setting thresholding and obtain court verdict, and by sentencing
Certainly result obtains sending the estimation of bit stream.
2. the OFDM method of reseptance of data model double drive according to claim 1, which is characterized in that right in the step 1
The loss function that deep neural network training uses is square error loss, and the optimizer used is that the estimation of adaptability momentum is excellent
Change device.
3. the OFDM method of reseptance of data model double drive according to claim 2, which is characterized in that adopted in the step 1
Loss function is square error loss, is specifically included:
In channel estimation module, the square error loss of use is:
In signal detection module, the square error loss of use is:
Wherein, N is sub-carrier number, and B is the bit number to be estimated, and H is actual frequency domain channel,It is channel estimation results, b
It is practical transmission bit stream,It is the estimation for sending bit stream.
4. the OFDM method of reseptance of data model double drive according to claim 1, which is characterized in that obtained in the step 2
Obtain LS channel estimation resultUsing formula:
Wherein, Y ' is received pilot tone, and X ' is local pilot tone.
5. the OFDM method of reseptance of data model double drive according to claim 1, which is characterized in that believe in the step 2
Deep neural network uses one layer of fully-connected network in road estimation module, whereinAlso, lowest mean square
Error signal estimationThe multiplying property parameter W of network is initialized as weight matrix WLMMSEReal value matrixWherein Re { } is to take real part, and Im { } is to take imaginary part；Additivity parameter n
When being initialized as complete zero, the initial value of network output isReal and imaginary parts series connection.
6. the OFDM method of reseptance of data model double drive according to claim 1, which is characterized in that in the step 3, obtain
Obtain zero forcing equalization resultUsing formula:
Wherein, Y is received frequency domain data,It is channel estimation results.
7. the OFDM method of reseptance of data model double drive according to claim 1, which is characterized in that deep in the step 3
Spend neural network specifically:
When ofdm system is linear system, using the deep neural network connected entirely；
When ofdm system is nonlinear system, using BiLSTM Recognition with Recurrent Neural Network and one layer of full connection depth nerve net
Network.
8. the OFDM method of reseptance of data model double drive according to claim 7, which is characterized in that in the step 3 when
It is equal to be estimated using the neuron number that a layer depth connects deep neural network full when ofdm system is nonlinear system
Bit number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810743534.1A CN109067688B (en) | 2018-07-09 | 2018-07-09 | Dual-drive OFDM (orthogonal frequency division multiplexing) receiving method of data model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810743534.1A CN109067688B (en) | 2018-07-09 | 2018-07-09 | Dual-drive OFDM (orthogonal frequency division multiplexing) receiving method of data model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109067688A true CN109067688A (en) | 2018-12-21 |
| CN109067688B CN109067688B (en) | 2021-09-07 |
Family
ID=64819570
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810743534.1A Active CN109067688B (en) | 2018-07-09 | 2018-07-09 | Dual-drive OFDM (orthogonal frequency division multiplexing) receiving method of data model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109067688B (en) |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109672639A (en) * | 2019-01-22 | 2019-04-23 | 哈尔滨工业大学 | A kind of signal demodulating method based on machine learning |
| CN109714086A (en) * | 2019-01-23 | 2019-05-03 | 上海大学 | Optimization MIMO detection method based on deep learning |
| CN109831399A (en) * | 2019-01-22 | 2019-05-31 | 东南大学 | A kind of OFDM receiver of artificial intelligence auxiliary |
| CN109831400A (en) * | 2019-03-06 | 2019-05-31 | 深兰科技（上海）有限公司 | A kind of received method and apparatus of OFDM signal |
| CN109889462A (en) * | 2019-01-21 | 2019-06-14 | 东南大学 | A kind of iteration receiving method of the neural network aiding suitable for high speed visible light communication |
| CN109921882A (en) * | 2019-02-20 | 2019-06-21 | 深圳市宝链人工智能科技有限公司 | A kind of MIMO coding/decoding method, device and storage medium based on deep learning |
| CN110266620A (en) * | 2019-07-08 | 2019-09-20 | 电子科技大学 | 3D MIMO-OFDM system channel estimation method based on convolutional neural networks |
| CN110661734A (en) * | 2019-09-20 | 2020-01-07 | 西安交通大学 | Channel estimation method, device and readable storage medium based on deep neural network |
| CN110719239A (en) * | 2019-09-29 | 2020-01-21 | 东南大学 | Data model dual-drive combined MIMO channel estimation and signal detection method |
| CN111245754A (en) * | 2020-01-03 | 2020-06-05 | 西安交通大学 | Self-adaptive space domain equalization method based on deep learning |
| CN111628946A (en) * | 2019-02-28 | 2020-09-04 | 华为技术有限公司 | Channel estimation method and receiving equipment |
| CN111669344A (en) * | 2020-06-01 | 2020-09-15 | 西北工业大学 | Time-varying OFDM system signal detection method based on deep learning |
| CN111786923A (en) * | 2020-06-05 | 2020-10-16 | 中国电子科技集团公司第七研究所 | Channel estimation method for time-frequency double-channel selection of orthogonal frequency division multiplexing system |
| CN112383496A (en) * | 2020-11-09 | 2021-02-19 | 西安科技大学 | Mine communication method, system, computer equipment and medium based on depth receiver |
| CN112637093A (en) * | 2020-12-09 | 2021-04-09 | 齐鲁工业大学 | Signal detection method based on model-driven deep learning |
| CN112866150A (en) * | 2021-01-13 | 2021-05-28 | 上海复旦通讯股份有限公司 | Wireless channel feature extraction and dimension reduction method and system |
| CN113037668A (en) * | 2021-05-20 | 2021-06-25 | 武汉科技大学 | Millimeter wave point-to-point communication channel equalization method |
| CN113676431A (en) * | 2021-07-08 | 2021-11-19 | 东南大学 | Model-driven MIMO-OFDM receiving method without cyclic prefix |
| CN113708891A (en) * | 2021-08-12 | 2021-11-26 | 电子科技大学 | Communication system adopting deep learning CS decoder |
| CN113839743A (en) * | 2021-09-18 | 2021-12-24 | 电子科技大学 | RLL code decoder for wireless communication receiving end |
| CN113872911A (en) * | 2021-10-15 | 2021-12-31 | 齐鲁工业大学 | Method and system for restraining peak-to-average power ratio of model-driven orthogonal frequency division multiplexing system |
| CN114301742A (en) * | 2021-12-23 | 2022-04-08 | 北京邮电大学 | Channel estimation method and device |
| CN114528925A (en) * | 2022-01-28 | 2022-05-24 | 华南理工大学 | Time-varying channel OFDM signal equalization method based on deep classification network |
| US20220239510A1 (en) * | 2021-01-25 | 2022-07-28 | Marvell Asia Pte Ltd | Ethernet physical layer transceiver with non-linear neural network equalizers |
| 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 |
| CN116455706A (en) * | 2023-04-12 | 2023-07-18 | 华南理工大学 | Channel estimation and equalization design method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101340266A (en) * | 2008-06-06 | 2009-01-07 | 山东大学 | Dual-selection channel tracking method based on particle filter and neural network |
| CN104022978A (en) * | 2014-06-18 | 2014-09-03 | 中国联合网络通信集团有限公司 | Half-blindness channel estimating method and system |
-
2018
- 2018-07-09 CN CN201810743534.1A patent/CN109067688B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101340266A (en) * | 2008-06-06 | 2009-01-07 | 山东大学 | Dual-selection channel tracking method based on particle filter and neural network |
| CN104022978A (en) * | 2014-06-18 | 2014-09-03 | 中国联合网络通信集团有限公司 | Half-blindness channel estimating method and system |
Non-Patent Citations (4)
| Title |
|---|
| ALEX GRAVES等: "Framewise Phoneme Classification with Bidirectional LSTM and Other Neural Network Architectures", 《NEURAL NETWORKS》 * |
| HAOYE等: "Power of Deep Learning for Channel Estimation and Signal Detection in OFDM Systems", 《IEEE WIRELESS COMMUNICATIONS LETTERS》 * |
| JUN SUN等: "Neural Network Channel Estimation Based on Least Mean Error Algorithm in the OFDM Systems", 《INTERNATIONAL SYMPOSIUM ON NEURAL NETWORKS:ADVANCES IN NEURAL NETWORKS》 * |
| TIM O’SHEA等: "An Introduction to Deep Learning for the Physical Layer", 《IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING》 * |
Cited By (39)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109889462A (en) * | 2019-01-21 | 2019-06-14 | 东南大学 | A kind of iteration receiving method of the neural network aiding suitable for high speed visible light communication |
| CN109889462B (en) * | 2019-01-21 | 2021-05-28 | 东南大学 | Neural network assisted iterative receiving method suitable for high-speed visible light communication |
| CN109831399A (en) * | 2019-01-22 | 2019-05-31 | 东南大学 | A kind of OFDM receiver of artificial intelligence auxiliary |
| CN109672639B (en) * | 2019-01-22 | 2021-07-16 | 哈尔滨工业大学 | Signal demodulation method based on machine learning |
| CN109672639A (en) * | 2019-01-22 | 2019-04-23 | 哈尔滨工业大学 | A kind of signal demodulating method based on machine learning |
| CN109714086A (en) * | 2019-01-23 | 2019-05-03 | 上海大学 | Optimization MIMO detection method based on deep learning |
| CN109714086B (en) * | 2019-01-23 | 2021-09-14 | 上海大学 | Optimized MIMO detection method based on deep learning |
| CN109921882A (en) * | 2019-02-20 | 2019-06-21 | 深圳市宝链人工智能科技有限公司 | A kind of MIMO coding/decoding method, device and storage medium based on deep learning |
| CN111628946A (en) * | 2019-02-28 | 2020-09-04 | 华为技术有限公司 | Channel estimation method and receiving equipment |
| CN111628946B (en) * | 2019-02-28 | 2021-10-26 | 华为技术有限公司 | Channel estimation method and receiving equipment |
| CN109831400A (en) * | 2019-03-06 | 2019-05-31 | 深兰科技（上海）有限公司 | A kind of received method and apparatus of OFDM signal |
| CN110266620A (en) * | 2019-07-08 | 2019-09-20 | 电子科技大学 | 3D MIMO-OFDM system channel estimation method based on convolutional neural networks |
| CN110661734B (en) * | 2019-09-20 | 2020-07-28 | 西安交通大学 | Channel estimation method, device and readable storage medium based on deep neural network |
| CN110661734A (en) * | 2019-09-20 | 2020-01-07 | 西安交通大学 | Channel estimation method, device and readable storage medium based on deep neural network |
| CN110719239A (en) * | 2019-09-29 | 2020-01-21 | 东南大学 | Data model dual-drive combined MIMO channel estimation and signal detection method |
| CN110719239B (en) * | 2019-09-29 | 2022-03-11 | 东南大学 | Data model dual-drive combined MIMO channel estimation and signal detection method |
| CN111245754A (en) * | 2020-01-03 | 2020-06-05 | 西安交通大学 | Self-adaptive space domain equalization method based on deep learning |
| CN111669344A (en) * | 2020-06-01 | 2020-09-15 | 西北工业大学 | Time-varying OFDM system signal detection method based on deep learning |
| CN111786923A (en) * | 2020-06-05 | 2020-10-16 | 中国电子科技集团公司第七研究所 | Channel estimation method for time-frequency double-channel selection of orthogonal frequency division multiplexing system |
| CN112383496A (en) * | 2020-11-09 | 2021-02-19 | 西安科技大学 | Mine communication method, system, computer equipment and medium based on depth receiver |
| CN112637093A (en) * | 2020-12-09 | 2021-04-09 | 齐鲁工业大学 | Signal detection method based on model-driven deep learning |
| CN112866150A (en) * | 2021-01-13 | 2021-05-28 | 上海复旦通讯股份有限公司 | Wireless channel feature extraction and dimension reduction method and system |
| CN112866150B (en) * | 2021-01-13 | 2024-05-17 | 上海复旦通讯股份有限公司 | Wireless channel feature extraction and dimension reduction method and system |
| US20220239510A1 (en) * | 2021-01-25 | 2022-07-28 | Marvell Asia Pte Ltd | Ethernet physical layer transceiver with non-linear neural network equalizers |
| CN113037668A (en) * | 2021-05-20 | 2021-06-25 | 武汉科技大学 | Millimeter wave point-to-point communication channel equalization method |
| CN113037668B (en) * | 2021-05-20 | 2023-03-10 | 武汉科技大学 | Millimeter wave point-to-point communication channel equalization method |
| CN113676431A (en) * | 2021-07-08 | 2021-11-19 | 东南大学 | Model-driven MIMO-OFDM receiving method without cyclic prefix |
| CN113708891A (en) * | 2021-08-12 | 2021-11-26 | 电子科技大学 | Communication system adopting deep learning CS decoder |
| CN113839743A (en) * | 2021-09-18 | 2021-12-24 | 电子科技大学 | RLL code decoder for wireless communication receiving end |
| CN113872911B (en) * | 2021-10-15 | 2023-10-24 | 齐鲁工业大学 | Model-driven method and system for suppressing peak-to-average ratio of orthogonal frequency division multiplexing system |
| CN113872911A (en) * | 2021-10-15 | 2021-12-31 | 齐鲁工业大学 | Method and system for restraining peak-to-average power ratio of model-driven orthogonal frequency division multiplexing system |
| CN114301742B (en) * | 2021-12-23 | 2024-01-09 | 北京邮电大学 | Channel estimation method and device |
| CN114301742A (en) * | 2021-12-23 | 2022-04-08 | 北京邮电大学 | Channel estimation method and device |
| CN114528925A (en) * | 2022-01-28 | 2022-05-24 | 华南理工大学 | Time-varying channel OFDM signal equalization method based on deep classification network |
| CN114528925B (en) * | 2022-01-28 | 2024-04-26 | 华南理工大学 | Time-varying channel OFDM signal equalization method based on deep classification network |
| 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 |
| CN116455706A (en) * | 2023-04-12 | 2023-07-18 | 华南理工大学 | Channel estimation and equalization design method and system |
| CN116455706B (en) * | 2023-04-12 | 2024-05-31 | 华南理工大学 | Channel estimation and equalization design method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109067688B (en) | 2021-09-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109067688A (en) | A kind of OFDM method of reseptance of data model double drive | |
| CN109617847B (en) | OFDM receiving method without cyclic prefix based on model-driven deep learning | |
| CN112600772B (en) | OFDM channel estimation and signal detection method based on data-driven neural network | |
| CN110163282B (en) | Modulation mode identification method based on deep learning | |
| CN109246038B (en) | Dual-drive GFDM receiver and method for data model | |
| CN112637093B (en) | Signal detection method based on model-driven deep learning | |
| Ma et al. | Signal demodulation with machine learning methods for physical layer visible light communications: Prototype platform, open dataset, and algorithms | |
| CN104811276B (en) | A kind of DL CNN demodulators of super Nyquist rate communication | |
| CN110266620A (en) | 3D MIMO-OFDM system channel estimation method based on convolutional neural networks | |
| Ha et al. | Machine learning model for adaptive modulation of multi-stream in MIMO-OFDM system | |
| CN113141325B (en) | Training method, identification method and device for optical OFDM signal subcarrier modulation format identification model | |
| Zhang et al. | Deep learning based on orthogonal approximate message passing for CP-free OFDM | |
| CN111565160B (en) | Combined channel classification, estimation and detection method for ocean communication system | |
| CN113285899A (en) | Time-varying channel estimation method and system based on deep learning | |
| CN111669344A (en) | Time-varying OFDM system signal detection method based on deep learning | |
| CN111431617A (en) | Asymmetric amplitude limiting light OFDM-IM modulation method for wireless optical communication | |
| CN114745230A (en) | OTFS signal receiving and recovering method based on deep neural network structure | |
| CN109861942A (en) | A kind of on-line study method of artificial intelligence auxiliary OFDM receiver | |
| Lv et al. | Joint OSNR monitoring and modulation format identification on signal amplitude histograms using convolutional neural network | |
| CN113708855A (en) | OTFS data drive receiving method, system and medium based on deep learning | |
| US20220405589A1 (en) | Methods for reliable over-the-air computation with pulses for distributed learning and with federated edge learning without channel state information | |
| CN109818889A (en) | Equalization algorithm for the SVM classifier optimization in high-order PAM optical transmission system | |
| Ye et al. | Bilinear convolutional auto-encoder based pilot-free end-to-end communication systems | |
| CN117640300A (en) | Channel estimation and demodulation method based on machine learning | |
| Yıldırım et al. | Deep receiver design for multi-carrier waveforms using cnns |
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 |