CN112380939A - Deep learning signal enhancement method based on generation countermeasure network - Google Patents
Deep learning signal enhancement method based on generation countermeasure network Download PDFInfo
- Publication number
- CN112380939A CN112380939A CN202011221709.6A CN202011221709A CN112380939A CN 112380939 A CN112380939 A CN 112380939A CN 202011221709 A CN202011221709 A CN 202011221709A CN 112380939 A CN112380939 A CN 112380939A
- Authority
- CN
- China
- Prior art keywords
- signal
- data
- noise
- generator
- target
- 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
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000013135 deep learning Methods 0.000 title claims abstract description 15
- 230000006870 function Effects 0.000 claims abstract description 25
- 238000010586 diagram Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 10
- 238000010606 normalization Methods 0.000 claims abstract description 3
- 230000004913 activation Effects 0.000 claims description 6
- 230000003042 antagnostic effect Effects 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims description 4
- 238000002474 experimental method Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Multimedia (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
- Noise Elimination (AREA)
Abstract
A deep learning signal enhancement method of generating a countermeasure network, comprising: extracting a high signal-to-noise ratio signal as a target signal, adding quantitative Gaussian noise into the target signal to obtain corresponding noise signal data with a low signal-to-noise ratio, and forming a data pair by the target signal data and the corresponding noise signal data to serve as an experimental data set; respectively defining model structures and loss functions of a generator and a discriminator, and carrying out normalization operation on data input into the discriminator; the data pair is used as the input of the discriminator, and the noise signal data in the data pair is used as the input of the generator. By countering the training, the difference between the generated data and the target data distribution is minimized, thereby obtaining a generator that achieves signal enhancement. The invention can adaptively learn the signal characteristics and realize signal enhancement. The invention improves the signal-to-noise ratio of the low signal-to-noise ratio signal, and intuitively obtains the denoised signal which is obviously cleaner than the noise signal by reconstructing the constellation diagram.
Description
Technical Field
The invention relates to a radio signal enhancement method.
Background
Noise is the most challenging problem for signals in wireless communications. The conventional solutions are linear and non-linear. Linear signal enhancement methods are widely used for noise reduction due to their relative simplicity, with the least mean square being the typical technique. However, the actual signal has non-stationary statistical characteristics, and the linear signal method has limited performance and cannot completely eliminate noise. Therefore, non-linear methods such as wavelet transform, non-linear threshold wavelet transform methods have been an active research area in the past few years because they can simultaneously clarify spectral and temporal information of signals or divide signals into different scale components by frequency range, filtering noise and interference. In some cases, however, such as at the point of discontinuity in the signal, the Gibbs phenomenon occurs after the signal has been enhanced. Moreover, if the threshold selection is incorrect, the signal enhancement effect is poor and the adaptability to noise fluctuation is poor.
For the above reasons, researchers in the field of wireless communication have also applied deep learning to the task of signal enhancement and achieved significant effects. For example, deep neural networks have been found to learn and analyze the noise and interference characteristics of wireless channels in addition to frequency selectivity. This shows that the deep learning method has certain advantages under the conditions of complex wireless channel and serious noise.
The patent application No. 200480024164.2 discloses a solution for signal reconstruction by using a generation countermeasure network, but only using a generator to change random noise into a signal similar to a real signal, and cannot realize the function of signal enhancement. The invention provides a method for realizing signal enhancement by a GAN-based signal enhancement network, which is different from other signal enhancement methods in that the other signal enhancement methods firstly extract the noise distribution of a signal and then eliminate the noise from the original signal. The invention discloses a robust method capable of adaptively learning signal characteristics and realizing time-varying system signal enhancement.
Disclosure of Invention
The present invention overcomes the above-mentioned shortcomings of the prior art, and provides a deep learning signal enhancement method based on generation of a countermeasure network.
The invention aims to provide a signal enhancement method based on GAN, which does not need to carry out complex preprocessing operation on original data, can adaptively extract the time sequence and the space characteristics of signals, realizes end-to-end signal enhancement, can effectively improve the signal-to-noise ratio of signals with low signal-to-noise ratio, and can more intuitively obtain the enhancement effect by reconstructing a constellation diagram. And through testing, the signal-to-noise ratio of the de-noised signal can be improved by 7 dB.
TABLE 1 SNR enhancement comparison of the present invention
| Number of test sets | Signal to noise ratio of noise signal | Signal to noise ratio of de-noised signal |
| 9000 | 10dB | 17.82dB |
| 3600 | 0dB | 7.93dB |
The technical scheme adopted by the invention for solving the technical problems is as follows:
a deep learning signal enhancement method based on a generation countermeasure network comprises the following steps:
s1: extracting high signal-to-noise ratio signals from the public data set as target signals, adding quantitative Gaussian noise into the signal data to obtain corresponding low signal-to-noise ratio noise signal data, and correspondingly forming data pairs;
s2: respectively defining model structures of a generator and a discriminator, and carrying out normalization operation on data input into the discriminator by adopting a Virtual Batch Norm algorithm to construct and generate a confrontation network model;
s3: respectively defining loss functions of a generator for generating an antagonistic network and an arbiter, taking a data pair as the input of the arbiter, taking noise data in the data pair as the input of the generator, and minimizing the difference between the generated data and the target data distribution through antagonistic training so as to obtain a generator for realizing signal enhancement;
s4: and extracting a trained generator, inputting a noise signal, and outputting a de-noising signal, wherein the evaluation method comprises the steps of calculating the signal-to-noise ratio of the de-noising signal and reconstructing a constellation diagram.
The technical idea of the invention is as follows: the training of the whole model framework is end-to-end, and the data input into the one-dimensional inverse convolution network in the generator is data with effective characteristics by combining random noise and the output data of the corresponding layer of the one-dimensional convolution network, so that the enhancement capability of the generator is improved. The data pairs are made, the additionally added noise data input into the discriminator are utilized to provide the characteristics of effective data in the noise signals, and the denoised signals generated by the generator can retain some data characteristics and remove other noise characteristics through the countermeasure between the discriminator and the generator, so that the enhancement effect is achieved. In addition, different training parameters such as different learning rates and different weight attenuations can be set, and the feasibility and the effectiveness of the method are proved by experimental results.
The invention has the beneficial effects that:
1) the deep convolutional network is used for automatically extracting the characteristics of the radio signals, the advantages of deep learning self-learning characteristics are fully utilized, the process of extracting noise is omitted, the defect of incomplete noise extraction is overcome, the complexity is greatly reduced, the process is simplified, and the time is saved;
2) the convolution network structure corresponding to 6 layers is used in the generator, the consistency of the data structure is ensured, the data input into the one-dimensional inverse convolution network is the data with effective characteristics combined with random noise and the output data of the corresponding layer of the one-dimensional convolution network, and the enhancement capability of the generator is improved. In addition, the data pairs are used as input in the discriminator, so that the characteristics of effective data in noise signals are provided, and the discrimination capability of the discriminator is improved;
3) the enhancement effect realized by the invention is very obvious in the reconstructed constellation diagram contrast diagram, and the signal-to-noise ratio of the noise data with low signal-to-noise ratio is obviously improved. Meanwhile, the noise removal of different low signal-to-noise ratio data can be realized by adjusting the training set, and even 0dB noise data with lower signal-to-noise ratio, which is not contained in the training set, can be effectively enhanced.
Drawings
FIG. 1 is a schematic diagram of the overall architecture of a system for carrying out the method of the present invention.
FIG. 2 is a diagram of a generator network model architecture of the present invention.
Fig. 3 is a diagram of a discriminator network model architecture of the present invention.
Fig. 4 is a comparison of reconstructed constellations of the present invention.
Detailed Description
The following detailed description of embodiments of the invention is provided in connection with the accompanying drawings.
Referring to fig. 1 to 4, a deep learning signal enhancement method based on a generative countermeasure network includes the following steps:
(1) constructing a data set required by an experiment:
step 11: a public data set RML2016.10a is adopted, data with the signal-to-noise ratio of 18dB and the modulation types of 8PSK, BPSK, CPFSK, GFSK, PAM4, QAM16, QAM64, QPSK and WBFM are extracted as target signals, and total 9000 data are used as main components in the data pairs which are provided for discriminators in the generation countermeasure network and are distinguished as real signals.
Step 12: and adding quantitative Gaussian noise into the extracted target signal data to obtain corresponding noise signal data with low signal-to-noise ratio. The target signal data and the low signal-to-noise ratio (SNR) obtained by expecting to add Gaussian noise are input, and the noise signal data with the corresponding low SNR can be obtained by the following formula calculation.
Nsignal＝Anoise+Asignal (5)
AnoiseTo normalize noise, RnoiseAs random noise, AsignalIs 18db signal amplitude, PsignalIs the signal frequency, n is the signal length, VnoiseAs variance of noise, σnoiseIs the standard deviation of noise, NsignalIs a noise signal.
Step 13: noise data of 15dB, 14dB, 13dB, 12dB, 11dB, 10dB, 9dB, 8dB and 3dB of low signal-to-noise ratio are obtained through the step 2, and target signal data and corresponding noise signal data form data pairs by utilizing an enum enumeration function and a zip function respectively to obtain a data set required by an experiment. 81000 sets of data pairs were included in the final experimental data set, with the data structure (None, 4, 128).
(2) Constructing and generating a confrontation network model:
step 21: defining a generator model, wherein L is the signal length and K is the number of channels. The first half of the network uses a 6-layer one-dimensional convolutional neural network, the convolutional kernel size is 3, the step size is 2, the padding is 1, and an activation function with parameters is used after each layer. And finally changing the input data structure from (None, 2, 128) to (None, 128, 2). The latter half network uses 6 layers of one-dimensional inverse convolution neural network, the convolution kernel size is 4, the step length is 2, the filling is 1, and an activation function with parameters is used after each layer; each layer of input is a combination of random noise (or the last layer of deconvolution output data) and the output data of the corresponding layer of the first half network. The final producer output data structure is still (None, 2, 128).
Step 22: defining a discriminator model, adopting a 6-layer convolution kernel with the size of 3 and the step length of 2, filling a one-dimensional convolution neural network with the size of 1, using a one-dimensional Virtual BatchNorm and a Leaky ReLU activation function after each layer, and adding a Dropout layer after the 3 rd layer convolution layer. And then 1 layer of one-dimensional convolutional neural network with the convolutional kernel size of 1 and the step length of 1, 1 linear layer and one sigmoid layer are connected. Finally, the data structure is changed from input (None, 4, 128) to output scalar.
(3) Training generates a confrontation network:
defining loss functions of a generator and a discriminator for generating an antagonistic network respectively: the discriminator mainly uses the least square error, which is the output value (f (x) of the discriminatori) Is minimized from the squared sum of the difference from the target value (1/0).
When data pairs of a target signal and a noise signal are input, the target value 1 is true, and the loss function is:
when the data pair of the denoised signal and the noise signal is input, the target value 0 is false, and the loss function is:
the generator combines the least square error and the least absolute deviation as a loss function, wherein the least absolute deviation is the de-noised signal data (G (x) generated by the generatori') and target signal data (S (x)i) ) is minimized.
The loss function of the generator adopts the two loss functions as follows:
and defining a model optimizer, inputting corresponding data pairs into a discriminator, and enabling the discriminator to discriminate whether the data pairs of the target signal and the noise signal are true or not and discriminate whether the data pairs of the de-noising signal and the noise signal are false or not. And inputting the noise signal data into a generator to generate a new denoising signal. And inputting the new data pair of the de-noised signal and the noise signal into a discriminator and expecting to judge the de-noised signal and the noise signal to be true so that the data of the de-noised signal and the target signal are more and more like. And finally, finishing training when the true and false discrimination probabilities of the discriminator on the data pairs are the same.
(4) And (3) saving and evaluating a denoised signal:
the trained generator is independently extracted, a noise signal is input, and a denoising signal is output; the evaluation method comprises the following steps of calculating the signal-to-noise ratio of a denoised signal and reconstructing a constellation diagram: subtracting the target signal from the de-noised signal to obtain a noise value, and meanwhile, taking the target signal as a signal value to obtain the signal-to-noise ratio of the de-noised signal by using a formula; and meanwhile, the difference between the de-noised signal and the noise signal is visually obtained by reconstructing the constellation diagram.
The embodiments described in this specification are merely illustrative of implementations of the inventive concept and the scope of the present invention should not be considered limited to the specific forms set forth in the embodiments but rather by the equivalents thereof as may occur to those skilled in the art upon consideration of the present inventive concept.
Claims (6)
1. A deep learning signal enhancement method based on a generation countermeasure network comprises the following steps:
s1: extracting high signal-to-noise ratio signals from the public data set as target signals, adding quantitative Gaussian noise into the signal data to obtain corresponding low signal-to-noise ratio noise signal data, and correspondingly forming data pairs;
s2: respectively defining model structures of a generator and a discriminator, and carrying out normalization operation on data input into the discriminator by adopting a Virtual Batch Norm algorithm to construct and generate a confrontation network model;
s3: respectively defining loss functions of a generator for generating an antagonistic network and an arbiter, taking a data pair as the input of the arbiter, taking noise data in the data pair as the input of the generator, and minimizing the difference between the generated data and the target data distribution through antagonistic training so as to obtain a generator for realizing signal enhancement;
s4: and extracting a trained generator, inputting a noise signal, and outputting a de-noising signal, wherein the evaluation method comprises the steps of calculating the signal-to-noise ratio of the de-noising signal and reconstructing a constellation diagram.
2. The deep learning signal enhancement method based on the generation countermeasure network as claimed in claim 1, wherein: step S1 specifically includes:
step 11: a public data set RML2016.10a is adopted, data with the signal-to-noise ratio of 18dB and the modulation types of 8PSK, BPSK, CPFSK, GFSK, PAM4, QAM16, QAM64, QPSK and WBFM are extracted as target signals, and total 9000 data are used as main components in the data pairs which are provided for discriminators in the generation countermeasure network and are distinguished as real signals.
Step 12: and adding quantitative Gaussian noise into the extracted target signal data to obtain corresponding noise signal data with low signal-to-noise ratio. The target signal data and the low signal-to-noise ratio (SNR) obtained by expecting to add Gaussian noise are input, and the noise signal data with the corresponding low SNR can be obtained by the following formula calculation.
Nsignal＝Anoise+Asignal (5)
AnoiseTo normalize noise, RnoiseAs random noise, AsignalIs 18db signal amplitude, PsignalIs the signal frequency, n is the signal length, VnoiseAs variance of noise, σnoiseIs the standard deviation of noise, NsignalIs a noise signal.
Step 13: noise data of 15dB, 14dB, 13dB, 12dB, 11dB, 10dB, 9dB, 8dB and 3dB of low signal-to-noise ratio are obtained through the step 2, and target signal data and corresponding noise signal data form data pairs by utilizing an enum enumeration function and a zip function respectively to obtain a data set required by an experiment.
3. The deep learning signal enhancement method based on the generation countermeasure network as claimed in claim 2, wherein: the data set of step 13 includes 81000 sets of data pairs, with the data structure (None, 4, 128).
4. The deep learning signal enhancement method based on the generation countermeasure network as claimed in claim 1, wherein: step S2 specifically includes:
step 21: defining a generator model, wherein L is the signal length and K is the number of channels. The first half of the network uses a 6-layer one-dimensional convolutional neural network, the convolutional kernel size is 3, the step size is 2, the padding is 1, and an activation function with parameters is used after each layer. Finally, the input data structure is changed from (None, 2, 128) to (None, 128, 2); the latter half network uses 6 layers of one-dimensional inverse convolution neural network, the convolution kernel size is 4, the step length is 2, the filling is 1, and an activation function with parameters is used after each layer; each layer of input is random noise or a combination of the last layer of deconvolution output data and the output data of the corresponding layer of the first half network; the data structure output by the final generator is still (None, 2, 128);
step 22: defining a discriminator model, adopting a 6-layer convolution kernel with the size of 3 and the step length of 2, filling a one-dimensional convolution neural network with the size of 1, using a one-dimensional Virtual Batch Norm and a Leaky ReLU activation function after each layer, and adding a Dropout layer after the 3 rd layer convolution layer. And then 1 layer of one-dimensional convolutional neural network with the convolutional kernel size of 1 and the step length of 1, 1 linear layer and one sigmoid layer are connected. Finally, the data structure is changed from input (None, 4, 128) to output scalar.
5. The deep learning signal enhancement method based on the generation countermeasure network as claimed in claim 1, wherein: step S3 specifically includes:
the discriminator uses the least square error to output the value (f (x)) of the discriminatori) Minimization of the sum of squares of the differences from the target value (1/0);
when data pairs of a target signal and a noise signal are input, the target value 1 is true, and the loss function is:
when the data pair of the denoised signal and the noise signal is input, the target value 0 is false, and the loss function is:
the generator combines the least square error and the least absolute deviation as a loss function, wherein the least absolute deviation is the de-noised signal data (G (x) generated by the generatori') and target signal data (S (x)i) Minimizing the sum of absolute differences;
the loss function of the generator adopts the two loss functions as follows:
defining a model optimizer, inputting corresponding data pairs into a discriminator, and enabling the discriminator to discriminate whether the data pairs of the target signal and the noise signal are true or not and discriminate whether the data pairs of the de-noising signal and the noise signal are false or not; inputting noise signal data into a generator to generate a new denoising signal, inputting the new denoising signal and the data pair of the noise signal into a discriminator, and expecting to judge the new denoising signal and the data pair of the noise signal to be true so that the data of the denoising signal and the target signal are more and more like; and finally, finishing training when the true and false discrimination probabilities of the discriminator on the data pairs are the same.
6. The deep learning signal enhancement method based on the generation countermeasure network as claimed in claim 1, wherein: step S4 specifically includes: the trained generator is independently extracted, a noise signal is input, and a denoising signal is output; the evaluation method comprises the following steps of calculating the signal-to-noise ratio of a denoised signal and reconstructing a constellation diagram: subtracting the target signal from the de-noised signal to obtain a noise value, and meanwhile, taking the target signal as a signal value to obtain the signal-to-noise ratio of the de-noised signal by using a formula; and meanwhile, the difference between the de-noised signal and the noise signal is visually obtained by reconstructing the constellation diagram.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011221709.6A CN112380939B (en) | 2020-11-05 | 2020-11-05 | Deep learning signal enhancement method based on generation countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011221709.6A CN112380939B (en) | 2020-11-05 | 2020-11-05 | Deep learning signal enhancement method based on generation countermeasure network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112380939A true CN112380939A (en) | 2021-02-19 |
| CN112380939B CN112380939B (en) | 2024-02-20 |
Family
ID=74578804
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011221709.6A Active CN112380939B (en) | 2020-11-05 | 2020-11-05 | Deep learning signal enhancement method based on generation countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112380939B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113111720A (en) * | 2021-03-17 | 2021-07-13 | 浙江工业大学 | Electromagnetic modulation signal denoising method and system based on deep learning |
| CN113537025A (en) * | 2021-07-08 | 2021-10-22 | 浙江工业大学 | Electromagnetic modulation signal deep denoising method and system based on self-supervision learning |
| CN113626960A (en) * | 2021-07-29 | 2021-11-09 | 南京信息工程大学 | Anti-mode coupling few-mode signal complex format analysis method and device |
| CN114091538A (en) * | 2021-11-23 | 2022-02-25 | 吉林大学 | Signal feature-based loss-discriminating convolutional neural network intelligent noise reduction method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110428849A (en) * | 2019-07-30 | 2019-11-08 | 珠海亿智电子科技有限公司 | A kind of sound enhancement method based on generation confrontation network |
| CN110619885A (en) * | 2019-08-15 | 2019-12-27 | 西北工业大学 | Method for generating confrontation network voice enhancement based on deep complete convolution neural network |
-
2020
- 2020-11-05 CN CN202011221709.6A patent/CN112380939B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110428849A (en) * | 2019-07-30 | 2019-11-08 | 珠海亿智电子科技有限公司 | A kind of sound enhancement method based on generation confrontation network |
| CN110619885A (en) * | 2019-08-15 | 2019-12-27 | 西北工业大学 | Method for generating confrontation network voice enhancement based on deep complete convolution neural network |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113111720A (en) * | 2021-03-17 | 2021-07-13 | 浙江工业大学 | Electromagnetic modulation signal denoising method and system based on deep learning |
| CN113111720B (en) * | 2021-03-17 | 2024-06-11 | 浙江工业大学 | Electromagnetic modulation signal denoising method and system based on deep learning |
| CN113537025A (en) * | 2021-07-08 | 2021-10-22 | 浙江工业大学 | Electromagnetic modulation signal deep denoising method and system based on self-supervision learning |
| CN113537025B (en) * | 2021-07-08 | 2024-03-22 | 浙江工业大学 | Electromagnetic modulation signal deep denoising method and system based on self-supervision learning |
| CN113626960A (en) * | 2021-07-29 | 2021-11-09 | 南京信息工程大学 | Anti-mode coupling few-mode signal complex format analysis method and device |
| CN113626960B (en) * | 2021-07-29 | 2022-06-14 | 南京信息工程大学 | Anti-mode coupling few-mode signal complex format analysis method and device |
| CN114091538A (en) * | 2021-11-23 | 2022-02-25 | 吉林大学 | Signal feature-based loss-discriminating convolutional neural network intelligent noise reduction method |
| CN114091538B (en) * | 2021-11-23 | 2024-04-30 | 吉林大学 | Intelligent noise reduction method for discrimination loss convolutional neural network based on signal characteristics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112380939B (en) | 2024-02-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112380939B (en) | Deep learning signal enhancement method based on generation countermeasure network | |
| CN110619885B (en) | Method for generating confrontation network voice enhancement based on deep complete convolution neural network | |
| CN110163282B (en) | Modulation mode identification method based on deep learning | |
| CN109890043B (en) | Wireless signal noise reduction method based on generative countermeasure network | |
| CN112464837B (en) | Shallow sea underwater acoustic communication signal modulation identification method and system based on small data samples | |
| CN108830809B (en) | Image denoising method based on expansion convolution | |
| CN109672639B (en) | Signal demodulation method based on machine learning | |
| CN110266620A (en) | 3D MIMO-OFDM system channel estimation method based on convolutional neural networks | |
| CN113723171B (en) | Electroencephalogram signal denoising method based on residual error generation countermeasure network | |
| CN112733811B (en) | Method for identifying underwater sound signal modulation modes based on improved dense neural network | |
| CN110490816B (en) | Underwater heterogeneous information data noise reduction method | |
| CN113572708B (en) | DFT channel estimation improvement method | |
| CN114422311B (en) | Signal modulation recognition method and system combining deep neural network and expert priori features | |
| CN110838089A (en) | Fast image denoising method based on OctBlock dense block | |
| CN113542171A (en) | Modulation pattern recognition method and system based on CNN and combined high-order spectral image | |
| CN110808057A (en) | Voice enhancement method for generating confrontation network based on constraint naive | |
| CN107094043B (en) | Improved MMSE low-complexity signal detection method based on block iteration method | |
| CN110430149B (en) | Least square channel estimation method based on minimum energy wavelet frame | |
| CN109167744B (en) | Phase noise joint estimation method | |
| CN115378776A (en) | MFSK modulation identification method based on cyclic spectrum parameters | |
| CN114301742B (en) | Channel estimation method and device | |
| ABOSINNEE et al. | Statistical vs. Information-theoretic signal properties over FFT-OFDM | |
| CN116016064B (en) | Communication signal noise reduction method based on U-shaped convolution denoising self-encoder | |
| CN114662523A (en) | Spread spectrum signal detection method integrating cognitive enhanced compressed sensing and artificial neural network | |
| CN115622849B (en) | Intelligent deceptive attack prevention method based on 5G signal MCS blind identification |
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 |