CN116074414A - Wireless communication physical layer structure based on deep learning - Google Patents

Wireless communication physical layer structure based on deep learning Download PDF

Info

Publication number
CN116074414A
CN116074414A CN202211556246.8A CN202211556246A CN116074414A CN 116074414 A CN116074414 A CN 116074414A CN 202211556246 A CN202211556246 A CN 202211556246A CN 116074414 A CN116074414 A CN 116074414A
Authority
CN
China
Prior art keywords
interference
encoder
decoder
output
subcarrier
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202211556246.8A
Other languages
Chinese (zh)
Inventor
金梦
朱丰源
解明奇
田晓华
王新兵
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.)
Shanghai Jiaotong University
Original Assignee
Shanghai Jiaotong 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 Shanghai Jiaotong University filed Critical Shanghai Jiaotong University
Priority to CN202211556246.8A priority Critical patent/CN116074414A/en
Publication of CN116074414A publication Critical patent/CN116074414A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/30—Definitions, standards or architectural aspects of layered protocol stacks
    • H04L69/32—Architecture of open systems interconnection [OSI] 7-layer type protocol stacks, e.g. the interfaces between the data link level and the physical level
    • H04L69/322—Intralayer communication protocols among peer entities or protocol data unit [PDU] definitions
    • H04L69/323—Intralayer communication protocols among peer entities or protocol data unit [PDU] definitions in the physical layer [OSI layer 1]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B1/00—Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission
    • H04B1/02—Transmitters
    • H04B1/04—Circuits
    • H04B1/0475—Circuits with means for limiting noise, interference or distortion
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/0082—Monitoring; Testing using service channels; using auxiliary channels
    • H04B17/0087—Monitoring; Testing using service channels; using auxiliary channels using auxiliary channels or channel simulators
    • 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
    • H04B17/345—Interference values
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/391—Modelling the propagation channel
    • H04B17/3912—Simulation models, e.g. distribution of spectral power density or received signal strength indicator [RSSI] for a given geographic region
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Computer Security & Cryptography (AREA)
  • Quality & Reliability (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Monitoring And Testing Of Transmission In General (AREA)

Abstract

The invention provides a wireless communication physical layer structure based on deep learning, which comprises the following steps: an encoder section, a channel simulation section, a decoder section, a random interference generator section, an interference feature extraction section, and a subcarrier restriction section. The invention realizes reliable communication under the severe condition of cross-protocol interference. The encoder end automatically adjusts the modulation mode of the signal according to the interference in the current environment, so that the encoder end has the capacity of resisting the interference, and the decoder of the self-encoder is used at the receiving end to restore the transmitted signal. The interference characteristic extractor and the subcarrier constrainer loaded at the encoder and decoder end are used for guiding the encoder to perform adaptive modulation mode control, so that different modulation modes are adopted for coping with different kinds of interference, and the effect of environmental adaptation is achieved. The design has strong universality, and can find a universal and effective anti-interference communication mode, so that the communication of a large number of Internet of things devices is guided, and support is provided for the development of the Internet of things field.

Description

Wireless communication physical layer structure based on deep learning
Technical Field
The invention relates to the technical field of communication technology, in particular to a wireless communication physical layer structure based on deep learning.
Background
In the age of the internet of things, the object of communication extends from a person to an object, which means that the diversity of network equipment is greatly increased, and the network equipment can be a smart phone or a bus card. Meanwhile, as application scenes are diversified, network environments become highly dynamic and complex, and communication modes are increasingly diversified.
In the patent document with publication number of CN113746628A, a physical layer key generation method and system based on deep learning are disclosed, by collecting the estimated value pairs of the channels of legal communication parties from both legal communication parties in the coherent time, the estimated value pairs respectively obtained by both legal communication parties are fused to obtain a pair of training data, a plurality of pairs of training data are obtained in a plurality of coherent time, a key generation network is established between both communication parties, and the key generation network comprises a feature extraction network and a decoding network; training a key generation network by using training data to realize network deep learning training, completing the training of the key generation network by sharing the pearson correlation coefficient and the mean value of each dimension of the consistent feature vector output by the feature extraction network of the two communication parties, generating the feature vector by using the trained key generation network according to the communication value, and quantizing the generated feature vector by using a key quantization algorithm to obtain a key sequence of the two communication parties.
Existing communication technologies are typically based on a more idealized and stiff network model, and it is difficult to achieve reliable transmission in highly dynamic and complex scenarios. Taking the physical layer of communication as an example, to implement wireless signal transmission, the primary task is to model a channel. As network environments become more complex, so too does the mathematical model used to describe the channel, from the initial free space transmission model to the final space-time channel model, the channel model needs to take into account path fading, shadowing, multipath effects, doppler phenomena, and the effects of multiple antennas on the channel at the same time. In addition to the interference between mass devices today, this makes the channel more difficult to estimate, and one cannot find an accurate mathematical model to describe and characterize the highly dynamic and complex channel. With the advent of the "ai+" age, this patent provided a channel modeling using a deep learning model (based on a self-encoder) instead of the traditional mathematical model. The method for adapting the channel model to the continuously-changing channel environment guides the Internet of things equipment to conduct high-reliability communication under the condition of interference.
Therefore, a new solution is needed to improve the above technical problems.
Disclosure of Invention
In view of the defects in the prior art, the invention aims to provide a wireless communication physical layer structure based on deep learning.
The invention provides a wireless communication physical layer structure based on deep learning, which comprises an encoder, a channel simulator, a decoder, a random interference generator, an interference feature extractor and a subcarrier constrainer, wherein the encoder is used for generating a random interference signal;
the encoder encodes and power constrains the signal according to the transmitted signal and the input interference characteristics;
the channel simulator superimposes the output of the encoder and the random interference generator in training to simulate the effect existing in a real channel;
the decoder decodes according to the interference characteristics of the output and input of the encoder through channel simulation;
the interference feature extractor performs feature extraction on the interference according to the output of the random interference generator through channel simulation;
the random interference generator is used as an interference source to simulate various interference possibly occurring in the environment;
the subcarrier constrainer constrains subcarriers output by the encoder according to the input interference information.
Preferably, the coding of the coder is based on a one-dimensional convolution layer and a full connection layer of a neural network, performs dimension transformation on an input signal, and modulates the signal on each subcarrier, wherein the output dimension is equal to the number of subcarriers;
the power constraint normalizes the encoded output and constrains the transmission power.
Preferably, the normalization refers to the maximum value of the signal amplitude on each subcarrier divided by the signal amplitude on that subcarrier.
Preferably, the effects in the channel simulator include carrier frequency offset CFO, sampling frequency offset SFO, gaussian noise, frequency conversion, pulse shaping.
Preferably, the decoding in the decoder performs a dimensional transformation on the input based on a one-dimensional convolution layer, a full connection layer and a regularization layer of the neural network, and outputs the transmitted signal.
Preferably, the feature extraction of the interference feature extractor comprises an interference feature extraction before the encoder and a feature extraction before the decoder;
the feature extraction is based on a one-dimensional convolution layer, a pooling layer and a full-connection layer of the neural network, performs dimension transformation on input interference, and searches for low-dimensional representation of an interference signal.
Preferably, the pre-encoder interference feature extractor uses the system error rate as constraint training, and the pre-decoder interference feature extractor uses the system error rate and the pre-encoder interference feature extractor output as constraint training;
the input of the interference feature extractor before the encoder and the input of the interference feature extractor before the decoder are staggered in time, and the features irrelevant to the interference and the time are extracted.
Preferably, the output of the interference feature extractor is connected to the encoder and decoder, directing the encoder and decoder to synchronize the encoder and decoder information.
Preferably, the subcarrier constraint is based on a one-dimensional convolution layer and a full connection layer of the neural network, and the output of the random interference generator through channel simulation is input, the subcarrier frequency band selection vector of the encoder is output, and the subcarrier frequency band selection vector is multiplied with the encoder output.
Preferably, the subcarrier band selection vector dimension is the same as the encoder output and only contains a 0/1 vector.
Compared with the prior art, the invention has the following beneficial effects:
through the design, the invention realizes reliable communication under the severe condition of cross-protocol interference; the method has strong universality, and can find a universal and effective anti-interference communication mode, so that the communication of a large number of Internet of things equipment is guided, and support is provided for the development of the Internet of things field.
Drawings
Other features, objects and advantages of the present invention will become more apparent upon reading of the detailed description of non-limiting embodiments, given with reference to the accompanying drawings in which:
FIG. 1 is a flow chart of the operation of the depth self-encoder based wireless communication physical layer architecture of the present invention;
FIG. 2 is a schematic diagram of the spectrum of interference generated by the random interference generator according to the present invention;
FIG. 3 is a diagram of simulated training effects;
FIG. 4 is an input interference constellation;
fig. 5 is a constellation diagram of the encoder output.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the present invention, but are not intended to limit the invention in any way. It should be noted that variations and modifications could be made by those skilled in the art without departing from the inventive concept. These are all within the scope of the present invention.
Example 1:
the invention provides a wireless communication physical layer structure based on deep learning, which comprises an encoder, a channel simulator, a decoder, a random interference generator, an interference feature extractor and a subcarrier constrainer, wherein the encoder is used for generating a random interference signal; the encoder encodes and power constrains the signal according to the transmitted signal and the input interference characteristics; the channel simulator superimposes the outputs of the encoder and the random interference generator in training to simulate the effect of the existence of a real channel; the decoder decodes according to the interference characteristics of the output and input of the encoder through channel simulation; the interference feature extractor performs feature extraction on the interference according to the output of the random interference generator through channel simulation; the random interference generator is used as an interference source to simulate various interference which can occur in the environment; the subcarrier constrainer constrains subcarriers output from the encoder according to the input interference information.
The coding of the coder is based on a one-dimensional convolution layer and a full connection layer of a neural network, performs dimension transformation on an input signal, outputs a dimension equal to the number of subcarriers, and modulates the signal on each subcarrier; the power constraint normalizes the encoded output and constrains the transmit power.
Normalization refers to dividing the signal amplitude on each subcarrier by the maximum of the signal amplitude on that subcarrier.
Effects in the channel simulator include carrier frequency offset CFO, sampling frequency offset SFO, gaussian noise, frequency conversion, pulse shaping.
Decoding in the decoder performs dimension transformation on the input based on a one-dimensional convolution layer, a full connection layer and a regularization layer of the neural network, and outputs the transmitted signal.
The feature extraction of the interference feature extractor comprises the feature extraction before the encoder and the feature extraction before the decoder; the feature extraction is based on a one-dimensional convolution layer, a pooling layer and a full-connection layer of the neural network, performs dimension transformation on input interference, and searches for low-dimensional representation of an interference signal.
The method comprises the steps that an interference feature extractor in front of an encoder takes a system error rate as constraint training, and the interference feature extractor in front of the decoder takes the system error rate and the output of the interference feature extractor in front of the encoder as constraint training; the input of the interference feature extractor before the encoder and the input of the interference feature extractor before the decoder are staggered in time, and features irrelevant to the interference are extracted.
The output of the interference feature extractor is connected with the encoder and the decoder, and guides the encoder and the decoder to synchronize the information of the encoder and the decoder.
The subcarrier constraint is based on a one-dimensional convolution layer and a full connection layer of the neural network, and is used for inputting the output of the random interference generator through channel simulation, outputting a subcarrier frequency band selection vector of the encoder and multiplying the subcarrier frequency band selection vector with the output of the encoder.
The subcarrier band select vector dimension is the same as the encoder output and contains only 0/1 of the vector.
Example 2:
according to the invention, the wireless communication physical layer structure based on deep learning comprises the following steps:
an encoder section, a channel simulation section, a decoder section, a random interference generator section, an interference feature extraction section, and a subcarrier restriction section.
The encoder section: encoding and power constraining the signal according to the transmitted signal and the input interference characteristics;
specifically, the encoding: based on the one-dimensional convolution layer and the full connection layer of the neural network, the dimension transformation is carried out on the input signal, the output dimension is equal to the number of subcarriers, and the modulation of the signal on each subcarrier is equivalent.
The power constraint: the encoded output is normalized to constrain the transmission power.
Specifically, the normalization refers to dividing the signal amplitude on each subcarrier by the maximum of the signal amplitude on that subcarrier.
The channel simulation section: superposition of the outputs of the encoder and the random disturbance generator during training simulates some effects of the real channel presence, including: carrier frequency offset CFO, sampling frequency offset SFO, gaussian noise, frequency conversion, pulse shaping.
The decoder section: decoding according to the interference characteristics of the output and input of the encoder through channel simulation;
specifically, the decoding: similar to the coding structure, the input is transformed in dimensions based on a one-dimensional convolution layer, a full connection layer and a regularization layer of the neural network, and the transmitted signals are output. Wherein the regularization layer functions to prevent overfitting of the training process.
The interference feature extractor portion: and extracting the characteristics of the interference according to the output of the random interference generator through channel simulation, wherein the characteristic extraction is divided into the characteristic extraction of the interference before the encoder and the characteristic extraction before the decoder.
Specifically, the feature extraction: based on a one-dimensional convolution layer, a pooling layer and a full-connection layer of the neural network, the input interference is subjected to dimension transformation, and the low-dimensional representation of the interference signal is found.
Specifically, the interference feature extraction before the encoder and the interference feature extraction before the decoder have the same network structure, and the difference is that:
the method comprises the steps that an interference feature extractor in front of an encoder takes a system error rate as constraint training, and the interference feature extractor in front of the decoder takes the system error rate and the output of the interference feature extractor in front of the encoder as constraint training;
the input of the pre-encoder interference feature extractor and the pre-decoder interference feature extractor are staggered in time in order to extract the interference-independent features.
Specifically, the output of the interference feature extractor is connected to the encoder and decoder for guiding the encoder and decoder while synchronizing the information of the encoder and decoder.
The random disturbance generator section: as an interference source, various kinds of interference which may occur in the simulation environment include:
Wi-Fi 802.11a/g；
Wi-Fi 802.11b/g；
Wi-Fi 802.11p；
BLE
LTE downlink RMC
LTE uplink RMC
OFDM
the subcarrier constraint part: and constraining the sub-carrier output by the encoder according to the input interference information so as to avoid the sub-carrier frequency band where the interference is located.
Specifically, the subcarrier constraints: based on the one-dimensional convolution layer and the full connection layer of the neural network, the output of the random interference generator through channel simulation is input, the subcarrier frequency band selection vector of the encoder is output, and the subcarrier frequency band selection vector is multiplied with the output of the encoder.
Specifically, the subcarrier frequency band selection vector: the dimensions are the same as the encoder output and only contain 0/1 of the vector.
In view of the defects in the prior art, the invention aims to provide a wireless communication physical layer structure based on deep learning.
The invention provides a wireless communication physical layer structure based on deep learning, which comprises the following parts:
an encoder section, a channel simulation section, a decoder section, a random interference generator section, an interference feature extraction section, and a subcarrier restriction section.
Part 1: an encoder:
the encoder performs dimension transformation and power constraint on the transmitted signal and the input interference characteristics based on a one-dimensional convolution layer and a full connection layer of the neural network, and the output dimension is equal to the number of subcarriers, which is equivalent to modulating the signal with interference on each subcarrier. Finally, power constraint is carried out: i.e. the encoded output is normalized, i.e. the signal amplitude on each subcarrier is divided by the maximum value of the signal amplitude on that subcarrier to constrain the transmission power.
Part 2: and (3) channel simulation:
the channel simulation part simulates some effects that may exist in a real channel in training and have an influence on communication quality, including carrier frequency offset CFO, sampling frequency offset SFO, gaussian noise, frequency conversion, and pulse shaping.
The specific process is as follows:
and (3) superposing sampling frequency offset SFO on the output (frequency domain) of the encoder, performing IFFT (inverse fast Fourier transform) to a time domain, performing up-sampling, pulse shaping and up-conversion, and superposing carrier frequency offset CFO and Gaussian noise to a certain extent. This is the output of channel A in FIG. 1.
And (4) carrying out FFT (fast Fourier transform) on 64-point output (time domain, corresponding to about 5 mu s of transmission time) of the random interference generator, then superposing sampling frequency offset SFO after the FFT is transferred to a frequency domain, then carrying out IFFT to transfer to the time domain, and superposing carrier frequency offset CFO. This is the output of channel C in FIG. 1.
Similarly, for 500-point output (time domain, corresponding to about 25 μs observation time) of the random interference generator, FFT is performed to the frequency domain and then the sampling frequency offset SFO is superimposed, and then IFFT is performed to the time domain and the carrier frequency offset CFO is superimposed. This is the output of channel B in FIG. 1.
And superposing the channel A and channel C outputs, and then performing down-conversion and down-sampling to serve as a first path of input of the decoder.
Part 3: a decoder:
similar to the coding structure, the decoder performs a dimensional transformation on the input based on a one-dimensional convolutional layer, a full-join layer, and a regularization layer of the neural network, and outputs the transmitted signal. Wherein the regularization layer functions to prevent overfitting of the training process.
The output of the decoder is the information finally demodulated by the receiving end.
Part 4: interference feature extractor:
an interference feature extractor section: and extracting the characteristics of the interference according to the output of the random interference generator through channel simulation, wherein the characteristic extraction is divided into the characteristic extraction of the interference before the encoder and the characteristic extraction before the decoder.
The feature extraction is based on a one-dimensional convolution layer, a pooling layer and a full-connection layer of the neural network, performs dimension transformation on input interference, and searches for low-dimensional representation of an interference signal.
The output dimension of the interference feature extractor should be less than or equal to the specific interference category number in the environment, so as to guide the encoder and the decoder to make the same modulation and demodulation reaction to the interference of the same general category.
Meanwhile, the interference feature extraction before the encoder and the interference feature extraction before the decoder have the same network structure, and the difference is that:
the method comprises the steps that an interference feature extractor in front of an encoder takes a system error rate as constraint training, and the interference feature extractor in front of the decoder takes the system error rate and the output of the interference feature extractor in front of the encoder as constraint training;
the input of the pre-encoder interference feature extractor and the pre-decoder interference feature extractor are staggered in time in order to extract the interference-independent features.
Specifically, the input to the interference feature extractor is the channel B output in fig. 1, i.e. the interference observations taking into account the real channel effects. The output of the interference feature extractor is the second input of the encoder and decoder.
Part 5: random interference generator:
random interference generator section: as an interference source, various kinds of interference which may occur in the simulation environment include:
Figure BDA0003983444660000071
part 6: subcarrier constraints:
and the subcarrier constraint is to constrain subcarriers output by the encoder according to the input interference information so as to avoid subcarrier frequency bands where the interference is located.
Specifically, based on a one-dimensional convolution layer and a full connection layer of the neural network, the output of the random interference generator through channel simulation is input, a subcarrier frequency band selection vector of the encoder is output, the vector dimension is the same as the output of the encoder, the vector dimension only comprises 0/1, whether the subcarrier is selected to transmit information or not is represented, and the subcarrier frequency band selection vector is multiplied with the output of the encoder.
The present embodiment will be understood by those skilled in the art as a more specific description of embodiment 1.
Those skilled in the art will appreciate that the invention provides a system and its individual devices, modules, units, etc. that can be implemented entirely by logic programming of method steps, in addition to being implemented as pure computer readable program code, in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Therefore, the system and various devices, modules and units thereof provided by the invention can be regarded as a hardware component, and the devices, modules and units for realizing various functions included in the system can also be regarded as structures in the hardware component; means, modules, and units for implementing the various functions may also be considered as either software modules for implementing the methods or structures within hardware components.
The foregoing describes specific embodiments of the present invention. It is to be understood that the invention is not limited to the particular embodiments described above, and that various changes or modifications may be made by those skilled in the art within the scope of the appended claims without affecting the spirit of the invention. The embodiments of the present application and features in the embodiments may be combined with each other arbitrarily without conflict.

Claims (10)

1. The wireless communication physical layer structure based on deep learning is characterized by comprising an encoder, a channel simulator, a decoder, a random interference generator, an interference feature extractor and a subcarrier constrainer;
the encoder encodes and power constrains the signal according to the transmitted signal and the input interference characteristics;
the channel simulator superimposes the output of the encoder and the random interference generator in training to simulate the effect existing in a real channel;
the decoder decodes according to the interference characteristics of the output and input of the encoder through channel simulation;
the interference feature extractor performs feature extraction on the interference according to the output of the random interference generator through channel simulation;
the random interference generator is used as an interference source to simulate various interference possibly occurring in the environment;
the subcarrier constrainer constrains subcarriers output by the encoder according to the input interference information.
2. The deep learning based wireless communication physical layer structure of claim 1, wherein the encoding of the encoder is based on a one-dimensional convolutional layer and a fully-connected layer of a neural network, performs a dimension transformation on an input signal, outputs a dimension equal to the number of subcarriers, and modulates the signal on each subcarrier;
the power constraint normalizes the encoded output and constrains the transmission power.
3. The deep learning based wireless communication physical layer structure of claim 2, wherein the normalization refers to a maximum value of the signal amplitude on each subcarrier divided by the signal amplitude on that subcarrier.
4. The deep learning based wireless communication physical layer structure of claim 1, wherein the effects in the channel simulator include carrier frequency offset CFO, sampling frequency offset SFO, gaussian noise, frequency conversion, pulse shaping.
5. The deep learning based wireless communication physical layer structure of claim 1, wherein the decoding in the decoder is based on a one-dimensional convolutional layer, a fully-connected layer, and a regularization layer of a neural network, performs a dimensional transformation on the input, and outputs the transmitted signal.
6. The deep learning based wireless communication physical layer structure of claim 1, wherein the feature extraction of the interference feature extractor comprises an interference feature extraction before encoder and a feature extraction before decoder;
the feature extraction is based on a one-dimensional convolution layer, a pooling layer and a full-connection layer of the neural network, performs dimension transformation on input interference, and searches for low-dimensional representation of an interference signal.
7. The deep learning based wireless communication physical layer structure of claim 6, wherein the pre-encoder interference feature extractor uses a system bit error rate as a constraint training, and the pre-decoder interference feature extractor uses the system bit error rate and an output of the pre-encoder interference feature extractor as a constraint training;
the input of the interference feature extractor before the encoder and the input of the interference feature extractor before the decoder are staggered in time, and the features irrelevant to the interference and the time are extracted.
8. The deep learning based wireless communication physical layer structure of claim 7, wherein the output of the interference feature extractor is coupled to an encoder and a decoder, directs the encoder and decoder to synchronize information of the encoder and decoder.
9. The deep learning based wireless communication physical layer structure of claim 1, wherein the subcarrier constraint is based on a one-dimensional convolution layer and a full connection layer of a neural network, inputs an output of a random interference generator through channel simulation, outputs a subcarrier frequency band selection vector of an encoder, and multiplies the encoder output.
10. The deep learning based wireless communication physical layer structure of claim 9, wherein the subcarrier band selection vector dimension is the same as the encoder output and contains only 0/1 vector.
CN202211556246.8A 2022-12-06 2022-12-06 Wireless communication physical layer structure based on deep learning Pending CN116074414A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211556246.8A CN116074414A (en) 2022-12-06 2022-12-06 Wireless communication physical layer structure based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211556246.8A CN116074414A (en) 2022-12-06 2022-12-06 Wireless communication physical layer structure based on deep learning

Publications (1)

Publication Number Publication Date
CN116074414A true CN116074414A (en) 2023-05-05

Family

ID=86177815

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211556246.8A Pending CN116074414A (en) 2022-12-06 2022-12-06 Wireless communication physical layer structure based on deep learning

Country Status (1)

Country Link
CN (1) CN116074414A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116401532A (en) * 2023-06-07 2023-07-07 山东大学 Method and system for recognizing frequency instability of power system after disturbance
CN117614784A (en) * 2023-11-15 2024-02-27 浙江恒业电子股份有限公司 Wireless communication module based on carrier wave
CN117614784B (en) * 2023-11-15 2024-06-07 浙江恒业电子股份有限公司 Wireless communication module based on carrier wave

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116401532A (en) * 2023-06-07 2023-07-07 山东大学 Method and system for recognizing frequency instability of power system after disturbance
CN116401532B (en) * 2023-06-07 2024-02-23 山东大学 Method and system for recognizing frequency instability of power system after disturbance
CN117614784A (en) * 2023-11-15 2024-02-27 浙江恒业电子股份有限公司 Wireless communication module based on carrier wave
CN117614784B (en) * 2023-11-15 2024-06-07 浙江恒业电子股份有限公司 Wireless communication module based on carrier wave

Similar Documents

Publication Publication Date Title
Ye et al. Power of deep learning for channel estimation and signal detection in OFDM systems
CN112737985B (en) Large-scale MIMO channel joint estimation and feedback method based on deep learning
CN113472706B (en) MIMO-OFDM system channel estimation method based on deep neural network
Ozpoyraz et al. Deep learning-aided 6G wireless networks: A comprehensive survey of revolutionary PHY architectures
Cai et al. Attention model for massive MIMO CSI compression feedback and recovery
CN106789764A (en) The transform domain quadratic estimate method of the denoising of joint Weighted Threshold and balanced judgement
Sun et al. ICINet: ICI-aware neural network based channel estimation for rapidly time-varying OFDM systems
CN116074414A (en) Wireless communication physical layer structure based on deep learning
CN109728824A (en) A kind of LDPC code iterative decoding method based on deep learning
Smith et al. A communication channel density estimating generative adversarial network
Lin et al. Modulation recognition using signal enhancement and multistage attention mechanism
Gao et al. Fully connected feedforward neural networks based CSI feedback algorithm
Zhao et al. Multi-task learning based underwater acoustic OFDM communications
CN114826832B (en) Channel estimation method, neural network training method, device and equipment
Liao et al. Deep learning channel estimation based on edge intelligence for NR-V2I
CN114124168A (en) MIMO-NOMA system signal detection method and system based on deep learning
CN116743220A (en) Robust symbol-level precoding method for resisting channel aging effect
Chen et al. Research on improved receiver of NOMA-OFDM signal based on deep learning
Zhao et al. Model-Driven Based Deep Unfolding Equalizer for Underwater Acoustic OFDM Communications
Zhang et al. Deep learning‐based digital signal modulation identification under different multipath channels
CN112671419B (en) Wireless signal reconstruction method, device, system, equipment and storage medium
CN115499276A (en) Channel estimation method, system and product of intelligent reflector auxiliary communication system
CN113242201B (en) Wireless signal enhanced demodulation method and system based on generation classification network
CN113852434A (en) LSTM and ResNet assisted deep learning end-to-end intelligent communication method and system
CN111769975A (en) MIMO system signal detection method and system

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