CN116800572A - QPSK receiver based on deep learning and auxiliary model training method thereof - Google Patents
QPSK receiver based on deep learning and auxiliary model training method thereof Download PDFInfo
- Publication number
- CN116800572A CN116800572A CN202310063191.5A CN202310063191A CN116800572A CN 116800572 A CN116800572 A CN 116800572A CN 202310063191 A CN202310063191 A CN 202310063191A CN 116800572 A CN116800572 A CN 116800572A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- model
- qpsk
- neural network
- training
- 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
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 64
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000012549 training Methods 0.000 title claims abstract description 41
- 238000003062 neural network model Methods 0.000 claims abstract description 38
- 230000006870 function Effects 0.000 claims abstract description 30
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 6
- 230000004913 activation Effects 0.000 claims description 14
- 210000002569 neuron Anatomy 0.000 claims description 14
- 238000001914 filtration Methods 0.000 claims description 5
- 230000006854 communication Effects 0.000 abstract description 16
- 238000004891 communication Methods 0.000 abstract description 15
- 238000013528 artificial neural network Methods 0.000 abstract description 13
- 230000000306 recurrent effect Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000010363 phase shift Effects 0.000 description 2
- 230000001105 regulatory effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/18—Phase-modulated carrier systems, i.e. using phase-shift keying
- H04L27/22—Demodulator circuits; Receiver circuits
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Error Detection And Correction (AREA)
Abstract
The application relates to a QPSK receiver auxiliary model training method based on deep learning, which comprises the following steps: building a QPSK system model, obtaining a training data set by using the QPSK system model, building a deep learning neural network model, generating a predicted value, training the deep learning neural network model by using the training data set, optimizing a loss function of the deep learning neural network model, optimizing the loss function by using a Nadam algorithm, updating parameters of the deep learning neural network model until the error rate of the predicted value compared with a true value is minimum, and taking the updated deep learning neural network model as an auxiliary model of the QPSK receiver. The model trained by the method adopts the deep learning neural network based on the dislike structure, and can be demodulated in the demodulation link of the communication receiver by considering the hidden relativity between values, and the problem that the recognition accuracy is not high due to noise interference under the condition of low signal-to-noise ratio in the traditional hard decision method can be solved by the demodulation of the model.
Description
Technical Field
The application belongs to the technical field of wireless communication, and particularly relates to a QPSK receiver based on deep learning and an auxiliary model training method thereof.
Background
In recent years, deep learning is widely applied to the fields of computer vision, unmanned driving, picture recognition and classification, target recognition and the like, and the reason why the deep learning is adopted is that the deep learning is an end-to-end method. Compared with other machine learning methods, the method has the advantages of wide coverage range, good adaptability, portability, very strong learning capacity and capability of learning deeper features from a large amount of data. In the field of wireless communication, combining deep learning with a wireless communication receiver is one of the hot directions of current academic research, and the defects of the conventional wireless communication receiver are likely to be repaired.
Quadrature phase shift keying (Quadrature Phase Shift Keying, QPSK) is widely used in wireless communication systems as a digital modulation scheme with high interference immunity, high spectrum utilization, and relatively simple circuit implementation. In addition to the channel noise interference, in the actual communication process, when two independent local oscillators are used on the transmitter and the receiver respectively and there is relative motion between the transmitter and the receiver, doppler shift occurs, which results in a certain frequency deviation between the received signal and the transmitted signal. Furthermore, due to imperfections of the radio frequency device, IQ imbalance may exist in the received IQ signal, i.e. an imbalance in the amplitude and/or phase of the I and Q channels. IQ imbalance can be described by a set of parameters (α, β), where α represents amplitude imbalance and β represents phase imbalance, which all affect the reception accuracy of the receiver. Under the condition of high signal-to-noise ratio, the traditional QPSK receiver frequently uses a hard decision method to demodulate from the received interference distortion signal, and then information is recovered through processes such as decoding, and the final error rate is high and the recognition accuracy is not high.
Therefore, a scheme is needed to solve the problem that the recognition accuracy is not high due to interference such as channel noise when the communication receiver receives the signal under the condition of high signal-to-noise ratio.
Disclosure of Invention
Based on the above-mentioned drawbacks and deficiencies of the prior art, it is an object of the present application to at least solve one or more of the above-mentioned problems of the prior art, in other words, to provide a deep learning based QPSK receiver and an auxiliary model training method thereof, which meet one or more of the above-mentioned needs.
In order to achieve the aim of the application, the application adopts the following technical scheme:
a QPSK receiver auxiliary model training method based on deep learning specifically comprises the following steps:
s1, constructing a QPSK system model, and acquiring a training data set by using the QPSK system model;
s2, constructing a deep learning neural network model, wherein the deep learning neural network model is used for demodulating the signals affected by the codes and noise to generate a predicted value;
s3, training the deep learning neural network model by using the training data set, and optimizing a loss function of the deep learning neural network model;
and S4, optimizing a loss function by using a Nadam algorithm, updating parameters of the deep learning neural network model until the error rate of comparison between the predicted value and the true value is minimum, and taking the updated deep learning neural network model as a QPSK receiver auxiliary model.
As a preferable scheme, a QPSK system model is used for acquiring a training data set, specifically, QPSK system model is used for carrying out QPSK modulation on data after Hamming coding, root raised cosine is used for carrying out oversampling, gaussian noise is added as interference, and 56-bit characteristics are obtained through matched filtering and undersampling.
As a preferred solution, the demodulation of the signal subjected to coding and noise influence by the deep learning neural network model specifically includes:
inputting a first full connection layer, wherein the first full connection layer uses a nonlinear activation function Relu;
inputting an LSTM layer, the LSTM layer having 128 neurons;
the input is output by a second fully connected layer having 56 neurons using an activation function sigmoid.
As a preferred solution, the loss function is a binary cross entropy.
In a second aspect, the present application provides a QPSK receiver-aided model training method based on a deep learning-aided model, which is a QPSK receiver-aided model trained by using any one of the above-mentioned deep learning-based QPSK receiver-aided model training methods, and is characterized by specifically comprising:
demodulating the signal subjected to coding and noise influence by using a QPSK receiver auxiliary model to generate a predicted value;
and carrying out hamming decoding on the predicted value, and recovering to obtain an original code.
As a preferred scheme, hamming decoding is (7, 4) decoding, outputting a 32-bit stream.
In a third aspect, the present application provides a QPSK intelligent receiver based on a deep learning auxiliary model, which uses the above-mentioned QPSK demodulation method based on a deep learning auxiliary model.
Compared with the prior art, the application has the beneficial effects that:
the model trained by the method adopts a deep learning neural network, and designs a neural network model by adopting a multi-label classification method, so as to replace a demodulation link of a communication receiver.
The main body part of the neural network model adopts an LSTM circulating neural network, and the memory structure in the recurrent neural network is used, so that the prediction performance is improved. The LSTM networks are all connected layers in front and behind, and are used for further improving the nonlinear expression capacity of the model so as to enhance the learning capacity of the model, so that the original information is better demodulated and recovered, and the problem that the identification accuracy is not high due to interference such as channel noise under the condition of high signal-to-noise ratio in the traditional hard decision method is solved.
Drawings
Fig. 1 is a flow chart of a deep learning based QPSK receiver aided model training method of the present application;
fig. 2 is a bit error rate comparison chart of demodulation by using the QPSK demodulation method according to the present application and demodulation by using the conventional hard decision method when the signal-to-noise ratio is 0-7 dB.
Fig. 3 is a bit error rate comparison chart of demodulation by using the QPSK demodulation method of the present application and demodulation by using the conventional hard decision method when the signal-to-noise ratio is 0 to 7dB and the normalized carrier frequency offset δf is set to 0.001, 0.002, and 0.004, respectively.
FIG. 4 is a bit error rate comparison graph of demodulation using the QPSK demodulation method of the application and demodulation using the conventional hard decision method when the signal-to-noise ratio is 0-7 dB, and three IQ imbalance configurations, (5, -6), (-3, 10) and (-3, -2);
fig. 5 is a schematic structural diagram of the deep learning-based QPSK communication intelligent receiver of the present application.
Detailed Description
In order to more clearly illustrate the embodiments of the present application, specific embodiments of the present application will be described below with reference to the accompanying drawings. It is evident that the drawings in the following description are only examples of the application, from which other drawings and other embodiments can be obtained by a person skilled in the art without inventive effort.
Examples: the application provides a QPSK receiver auxiliary model training method based on deep learning, wherein a flow chart of the method is shown in figure 1, and in a certain embodiment of the application, the method specifically comprises the following steps:
s1, constructing a QPSK system model, and acquiring a training data set by using the QPSK system model;
s2, constructing a deep learning neural network model, wherein the deep learning neural network model is used for demodulating the signals affected by the codes and noise to generate a predicted value;
s3, training the deep learning neural network model by using the training data set, and optimizing a loss function of the deep learning neural network model;
and S4, optimizing a loss function by using a Nadam algorithm, updating parameters of the deep learning neural network model until the error rate of comparison between the predicted value and the true value is minimum, and taking the updated deep learning neural network model as a QPSK receiver auxiliary model.
In a further embodiment of step S1, the training data set is obtained by using a QPSK system model, specifically, the data after hamming encoding is QPSK modulated by using a QPSK system model, then oversampled by using root raised cosine, and gaussian noise is added as interference, and then the 56-bit feature is obtained through matched filtering and undersampling.
One piece of data to be transmitted by the QPSK system model is 01 bit stream, and the data is randomly generated by MATLAB and has 32 bits in total. In a QPSK system model, carrying out (7, 4) Hamming coding on data to be transmitted at a transmitting end to obtain 56 bits, carrying out QPSK modulation, carrying out oversampling by utilizing root raised cosine, wherein the number of samples per symbol is 8, adding Gaussian noise as interference, and carrying out matched filtering and undersampling: every 8 samples take one symbol, then a 56-bit feature is obtained, and a 56-bit hamming code obtained by hamming coding is used as a value. All features generated through the above-described QPSK system model together with corresponding values constitute a training data set.
In a further embodiment of step S2, the demodulation of the encoded and noise affected signal by the deep learning neural network model specifically comprises:
inputting a first full connection layer, wherein the first full connection layer uses a nonlinear activation function Relu;
inputting an LSTM layer, the LSTM layer having 128 neurons;
the input is output by a second fully connected layer having 56 neurons using an activation function sigmoid.
As a further preferred embodiment, the predicted value is 56 bits.
Specifically, the first layer of the deep learning neural network is a full-connection layer, 128 neurons are all used, the adopted activation function is a nonlinear activation function Relu, the second layer is a main body part of the network, LSTM is adopted, 128 neurons are all used, the adopted activation function is Relu, the memory structure in the recurrent neural network is utilized, so that the prediction performance is improved, the last layer is the full-connection layer, 56 neurons are all used as an output layer, the adopted activation function is sigmoid, the output result of the LSTM network is input into the sigmoid layer for calculation, each real number in an input vector is mapped into a real number between 0 and 1, and all values in the output vector are in a range of [0,1], so that the multi-label classification method is realized.
The steps S3 and S4 use the deep learning neural network to output the result of five iterations, take the result of the fifth iteration as the output result, be the 56-bit floating point number, obtain the 56-bit predicted value (pre_label) after 01 judgment, compare with the true value (true_label), calculate and get the error rate, thus calculate the loss function of the deep learning neural network model according to the output result of the neural network.
And then optimizing the loss function by adopting a Nadam algorithm, carrying out back propagation training on the fully connected neural network model, and updating and optimizing the parameter values and weights of the neurons of each layer.
And in the updating process, the model parameters are regulated, so that a predicted value (pre_label) obtained by carrying out 01 judgment on the output information is compared with a true value (true_label), the calculated error rate is minimum, and the training of the deep learning neural network model is completed.
Specifically, the loss function is set to binary cross entropy. The definition is as follows:
where the batch size is N, y is binary tag 0 or 1, and p (y) is the probability of the output belonging to y tags.
The following provides a specific implementation procedure in a certain embodiment of the present application:
in this embodiment, it is specified in step S1 that the signal-to-noise ratio ranges from 0 to 7dB for each training data set, the interval is 1dB, and the number of data samples for each signal-to-noise ratio is 20 ten thousand, so the total training sample size is 160 ten thousand (20 ten thousand×8). For each test dataset, the signal to noise ratio ranged from 0dB to 7dB, with 0.5dB spacing, and the number of samples per signal to noise ratio was 10 tens of thousands, so the total test sample size was 150 tens of thousands (10 tens of thousands 15). One piece of data to be transmitted in step S1 is a 01 bit stream, and the information bits are 4 bits, and 8 groups are total, that is, 32 bits. The data to be transmitted is (7, 4) Hamming coded at the transmitting end to obtain 56 bits, then QPSK modulated, and the root raised cosine is utilized to carry out oversampling, the sampling number of each symbol is 8, then Gaussian noise is added as interference, and the matching filtering and undersampling are carried out: every 8 samples take one symbol, then a 56-bit feature is obtained, and a 56-bit hamming code obtained by hamming coding is used as a value. All features together with the corresponding values constitute a dataset.
In step S2, the first layer of the deep learning network is a fully connected layer, there are 128 neurons in total, the adopted activation function is a nonlinear activation function Relu, the second layer is a main body part of the network, LSTM is adopted, there are 128 neurons in total, the adopted activation function is Relu, the memory structure in the recurrent neural network is used, so as to improve the prediction performance, the last layer is a fully connected layer, there are 56 neurons in total, the adopted activation function is sigmoid, the output result of the LSTM network is input to the sigmoid layer for calculation, so that each real number in the input vector is mapped into a real number between 0 and 1, and all values in the output vector are in the interval of [0,1 ].
And S3 and S4, using a 56-bit floating point number result output by the deep learning neural network, carrying out 01 judgment to obtain a 56-bit predicted value (pre_label), comparing the 56-bit predicted value (pre_label) with a true value (true_label), and calculating to obtain an error rate, thereby calculating a loss function of the deep learning neural network model according to the output result of the neural network.
And then optimizing the loss function by adopting a Nadam algorithm, carrying out back propagation training on the fully connected neural network model, and updating and optimizing the parameter values and weights of the neurons of each layer.
The loss function is set to binary cross entropy. The definition is as follows:
where the batch size is N, y is binary tag 0 or 1, and p (y) is the probability of the output belonging to y tags.
And in the updating process, the model parameters are regulated, so that a predicted value (pre_label) obtained by carrying out 01 judgment on the output information is compared with a true value (true_label), the calculated error rate is minimum, and the training of the deep learning neural network model is completed.
The model trained by the method is based on the deep learning QPSK communication intelligent receiver, and a neural network model is designed by adopting a multi-label classification method and is used for replacing the demodulation link of the communication receiver. The main body part of the neural network model adopts LSTM, and the memory structure in the recurrent neural network is used, so that the prediction performance is improved. The LSTM networks are all connected layers in front and behind, and are used for further improving the nonlinear expression capacity of the model so as to enhance the learning capacity of the model. And sending the interference distortion signals transmitted by the channels into a QPSK communication intelligent receiver for demodulation, and recovering the original information by Hamming decoding. The method solves the problem that the identification accuracy is not high due to interference such as channel noise and the like under the condition of high signal-to-noise ratio in the traditional hard decision method.
In a second aspect, the present application provides a QPSK receiver-aided model training method based on a deep learning-aided model, which is a QPSK receiver-aided model trained by using any one of the above-mentioned deep learning-based QPSK receiver-aided model training methods, and is characterized by specifically comprising:
demodulating the signal subjected to coding and noise influence by using a QPSK receiver auxiliary model to generate a predicted value;
and carrying out hamming decoding on the predicted value, and recovering to obtain an original code.
In a preferred embodiment, the hamming decoding is a (7, 4) decoding, outputting a 32-bit bitstream. Specifically, (7, 4) hamming decoding is performed on the 56-bit predicted value (pre_label) to obtain a 32-bit stream, namely, the restored information bit. And comparing the restored information bit with the original information bit, and calculating the bit error rate for determining the accuracy of the QPSK demodulation method.
The implementation uses MATLAB R2022b and pyr 2020, tensorf low2.4.0CPU to simulate the steps to carry out simulation experiments, calculates and draws a bit error rate comparison chart calculated by demodulating by using the QPSK demodulation method and demodulating by using the traditional hard decision method along with the increase of the signal to noise ratio when the signal to noise ratio is 0-7 dB, as shown in figure 2. The normalized carrier frequency offset δf (with respect to the symbol rate) is set to 0.001, 0.002, 0.004, respectively. The bit error rate calculated by demodulating by using the QPSK demodulation method and demodulating by using the traditional hard decision method is compared with the figure 3. Three IQ imbalance configurations: and (5, -6), (-3, 10) and (-3, -2) using the QPSK demodulation method of the present application and the conventional hard decision method, as shown in fig. 4.
In a third aspect, the present application provides a QPSK intelligent receiver based on a deep learning auxiliary model, which uses the above-mentioned QPSK demodulation method based on a deep learning auxiliary model.
The receiver adopts a QPSK communication intelligent receiver based on deep learning at a receiving end, and designs a neural network model by adopting a multi-label classification method, so as to replace a demodulation link of the communication receiver. The main body part of the neural network model adopts LSTM, and the memory structure in the recurrent neural network is used, so that the prediction performance is improved. The LSTM networks are all connected layers in front and behind, and are used for further improving the nonlinear expression capacity of the model so as to enhance the learning capacity of the model. And sending the interference distortion signals transmitted by the channels into a QPSK communication intelligent receiver for demodulation, and recovering the original information by Hamming decoding. The method solves the problem that the identification accuracy is not high due to interference such as channel noise and the like under the condition of high signal-to-noise ratio in the traditional hard decision method.
It is to be understood that the foregoing is only illustrative of the preferred embodiments and concepts of the application and that modifications in this detailed description will readily suggest themselves to those skilled in the art in view of the teachings of this application, and are to be regarded as illustrative of the scope of the application.
Claims (8)
1. The QPSK receiver auxiliary model training method based on deep learning is characterized by comprising the following steps of:
s1, constructing a QPSK system model, and acquiring a training data set by using the QPSK system model;
s2, constructing a deep learning neural network model, wherein the deep learning neural network model is used for demodulating the signals affected by the codes and noise to generate predicted values;
s3, training the deep learning neural network model by using the training data set, and optimizing a loss function of the deep learning neural network model;
and S4, optimizing the loss function by using a Nadam algorithm, updating parameters of the deep learning neural network model until the error rate of comparison between the predicted value and the true value is minimum, and taking the updated deep learning neural network model as a QPSK receiver auxiliary model.
2. The method for training a deep learning-based QPSK receiver according to claim 1, wherein the training data set is obtained by using the QPSK system model, specifically, the data after hamming encoding is QPSK modulated by using the QPSK system model, then oversampled by using root raised cosine, and then gaussian noise is added as channel interference, and 56-bit features are obtained by matching filtering and undersampling.
3. The method for training a deep learning-based QPSK receiver assist model as recited in claim 1, wherein the demodulation of the encoded and noise-affected signal by the deep learning neural network model specifically includes:
inputting a first fully connected layer, wherein the first fully connected layer uses a nonlinear activation function Relu;
inputting an LSTM layer, the LSTM layer having 128 neurons;
the second fully connected layer, which has 56 neurons, is input for output using the activation function sigmoid.
4. A QPSK receiver assist model training method as defined in claim 3, wherein said predicted value is 56 bits.
5. The method for training a deep learning based QPSK receiver assist model as recited in claim 1, wherein the loss function is a binary cross entropy.
6. A QPSK receiver-aided model training method based on a deep learning-aided model, which is a QPSK receiver-aided model trained by using the QPSK receiver-aided model training method based on deep learning according to any one of claims 1 to 5, comprising:
demodulating the signals subjected to coding and noise influence by using the QPSK receiver auxiliary model to generate a predicted value;
and carrying out hamming decoding on the predicted value, and restoring to obtain an original code.
7. The QPSK demodulation method as recited in claim 6, wherein the hamming decoding is a (7, 4) decoding, outputting a 32-bit bitstream.
8. A QPSK intelligent receiver based on a deep learning assist model, wherein a QPSK demodulation method based on a deep learning assist model according to any of claims 6-7 is used.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310063191.5A CN116800572A (en) | 2023-01-19 | 2023-01-19 | QPSK receiver based on deep learning and auxiliary model training method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310063191.5A CN116800572A (en) | 2023-01-19 | 2023-01-19 | QPSK receiver based on deep learning and auxiliary model training method thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116800572A true CN116800572A (en) | 2023-09-22 |
Family
ID=88040557
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310063191.5A Pending CN116800572A (en) | 2023-01-19 | 2023-01-19 | QPSK receiver based on deep learning and auxiliary model training method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116800572A (en) |
-
2023
- 2023-01-19 CN CN202310063191.5A patent/CN116800572A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110166391B (en) | Baseband precoding MSK signal demodulation method based on deep learning under impulse noise | |
| Tang et al. | A new low-complexity near-ML detection algorithm for spatial modulation | |
| Wu et al. | Deep learning-driven wireless communication for edge-cloud computing: opportunities and challenges | |
| CN113630130B (en) | End-to-end digital communication demodulation method | |
| CN106911374A (en) | A kind of low complexity soft exports spatial modulation detection method | |
| CN110233810B (en) | MSK signal demodulation method based on deep learning under mixed noise | |
| Lin et al. | A real-time modulation recognition system based on software-defined radio and multi-skip residual neural network | |
| Zheng et al. | DemodNet: Learning soft demodulation from hard information using convolutional neural network | |
| CN101529846A (en) | QAM receiver | |
| Meenalakshmi et al. | Enhancing channel estimation accuracy in polar-coded MIMO–OFDM systems via CNN with 5G channel models | |
| CN113612720B (en) | Anti-interference modulation method and system for wireless channel | |
| CN113114603B (en) | Information recovery method and device for MIMO-OFDM system | |
| Qing et al. | ELM-based frame synchronization in nonlinear distortion scenario using superimposed training | |
| CN115250217B (en) | OFDM system channel estimation method and device assisted by differential coding and neural network | |
| CN116800572A (en) | QPSK receiver based on deep learning and auxiliary model training method thereof | |
| Shevitski et al. | Digital signal processing using deep neural networks | |
| CN116015536A (en) | Incoherent reflection modulation method | |
| CN109245854A (en) | A kind of end-to-end wireless communication system and method based on AI | |
| Che et al. | A low-complexity transceiver design for terahertz communication based on deep learning | |
| WO2009035495A2 (en) | Optimal coherent demodulation for d-psk | |
| Jeon et al. | Reinforcement-learning-aided detector for time-varying MIMO systems with one-bit ADCs | |
| Zhao et al. | An End-to-End Demodulation System Based on Convolutional Neural Networks | |
| CN113726702A (en) | Low-complexity detection method for ubiquitous sensing network coding MPSK signal | |
| Wang et al. | A Signal processing method of OFDM communication receiver based on CNN | |
| CN116055273A (en) | QPSK receiver cascaded by neural network and auxiliary model training method thereof |
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 |