CN113572710A - WVD time-frequency analysis cross item suppression method and system based on generation countermeasure network and storage medium - Google Patents

WVD time-frequency analysis cross item suppression method and system based on generation countermeasure network and storage medium Download PDF

Info

Publication number
CN113572710A
CN113572710A CN202110826554.7A CN202110826554A CN113572710A CN 113572710 A CN113572710 A CN 113572710A CN 202110826554 A CN202110826554 A CN 202110826554A CN 113572710 A CN113572710 A CN 113572710A
Authority
CN
China
Prior art keywords
cross
discriminator
generator
wvd
network
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
CN202110826554.7A
Other languages
Chinese (zh)
Inventor
谭龙飞
崔梦玲
叶鑫
钱江
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.)
Sichuan Fire Research Institute of Emergency Management Department
Yangtze River Delta Research Institute of UESTC Huzhou
Original Assignee
Sichuan Fire Research Institute of Emergency Management Department
Yangtze River Delta Research Institute of UESTC Huzhou
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 Sichuan Fire Research Institute of Emergency Management Department, Yangtze River Delta Research Institute of UESTC Huzhou filed Critical Sichuan Fire Research Institute of Emergency Management Department
Priority to CN202110826554.7A priority Critical patent/CN113572710A/en
Publication of CN113572710A publication Critical patent/CN113572710A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
    • H04L25/03006—Arrangements for removing intersymbol interference
    • H04L25/03159—Arrangements for removing intersymbol interference operating in the frequency domain
    • 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
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
    • H04L25/03006—Arrangements for removing intersymbol interference
    • H04L25/03012—Arrangements for removing intersymbol interference operating in the time domain
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
    • H04L25/03006—Arrangements for removing intersymbol interference
    • H04L25/03178—Arrangements involving sequence estimation techniques
    • H04L25/03248—Arrangements for operating in conjunction with other apparatus
    • H04L25/0328—Arrangements for operating in conjunction with other apparatus with interference cancellation circuitry

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Power Engineering (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a WVD time-frequency analysis cross item suppression method, a system and a storage medium based on a generated countermeasure network, wherein a generated countermeasure GAN network is firstly constructed, and comprises a generator G and a discriminator D: the generator G is used for acquiring sample data; the discriminator D is used for judging whether the input data is real data; acquiring an image containing a cross item and an image for eliminating the cross item; the generator and the discriminator are trained in a mutual game until the cross terms in the image containing the cross terms are eliminated. The method provided by the invention combines a deep learning network, and utilizes the end-to-end characteristic of the network and the convenience and diversity of the preparation of cross item data; the cross terms generated by the multi-component frequency modulation signal in the WVD conversion are eliminated. Is more rapid than the conventional method. Under the condition of high signal-to-noise ratio, the method can be continuously exerted only by changing the training set, and is more accurate than the cross item elimination used by the traditional method.

Description

WVD time-frequency analysis cross item suppression method and system based on generation countermeasure network and storage medium
Technical Field
The invention relates to the technical field of target detection, in particular to a WVD time-frequency analysis cross item suppression method and system based on a generation countermeasure network.
Background
Although the frequency resolution of the WVD (Wigner-Ville distribution) method is very high, cross terms are generated, which seriously affects the quality of the ISAR image, and how to suppress the cross terms in the WVD (Wigner-Ville distribution) transform is a hot spot in the field of signal processing.
Currently, there are comparative methods (e.g., smooth pseudo-Wigner distribution, Choi Williams distribution, etc.) that can effectively suppress cross-term interference. The aim of removing cross terms is achieved by decomposing multi-component signals into single-component signals by using Fourier Bessel expansion of Ram Bilas Pachoi and the like, calculating WVD (WVD) for each single-component signal and combining the single-component signals. The Chen end and the like inhibit cross terms based on Gabor transformation, but the method calculates WVD by using Gabor coefficients after decomposing signals, has the defects that the composition components of the signals cannot be observed visually, and the signals can be reconstructed only by using fixed Gabor coefficients, and can generate larger aliasing and fail when each component signal is relatively close.
Thus, conventional methods can suppress some of the cross terms to some extent, but at the expense of losing their resolution. Particularly when the component signals are relatively close, aliasing occurs and the conventional method fails.
Disclosure of Invention
In view of the above, the present invention provides a method for suppressing cross terms in WVD time-frequency analysis based on a generative countermeasure network, which can accurately eliminate cross terms.
In order to achieve the purpose, the invention provides the following technical scheme:
the invention provides a WVD time-frequency analysis cross item suppression method based on a generation countermeasure network, which comprises the following steps:
constructing a generative antagonistic GAN network, said GAN network comprising a generator G and an arbiter D: the generator G is used for acquiring sample data; the discriminator D is used for judging whether the input data is real data;
acquiring an image containing a cross item and an image for eliminating the cross item;
the generator and the discriminator are trained in a mutual game until the cross terms in the image containing the cross terms are eliminated.
Further, the training of the generator and the discriminator in a mutual game mode comprises the following specific steps:
training parameters of a discriminator D:
generating a sample G (z) by inputting the image containing the cross terms into a generator G;
inputting the sample G (z) and the image for eliminating the cross terms into a discriminator D for training to obtain parameters of the discriminator D;
outputting the discrimination result of the discriminator D;
parameter training of the generator:
adjusting the parameters of the generator according to the judgment result output by the judger D until the output result of the generator G is in a real state;
the parameters of generator G are retained.
Further, the generative antagonistic GAN network employs a pix2pix network.
Further, the generator G (x) adopts a U-Net network and comprises a down-sampling coding unit, an up-sampling coding unit and a connecting unit;
the U-Net network is carried out according to the following steps:
down-sampling an input image to a low-dimensional image through a sampling coding unit, and up-sampling the low-dimensional image to an original resolution through an up-sampling coding unit;
and splicing the down sampling and the up sampling according to corresponding channels through a connecting unit.
Further, the discriminator adopts Markov discriminator PatchGAN.
Further, the full convolution in the Markov discriminator PatchGAN adopts a full convolution small network, the probability that the sigmoid output of each pixel in the last layer is true is output, and the final loss is obtained by BCEloss calculation.
Further, the image containing the cross terms is manufactured according to the following steps:
by applying a linear frequency-modulated signal s1(t) and s2(t) adding to obtain s (t), and then carrying out WVD conversion on s (t) to obtain data W containing cross termss(t,w)。
Further, the cross term elimination image is manufactured according to the following steps:
separately applying a linear frequency-modulated signal s1(t) and s2(t) performing a WVD conversion to obtain
Figure BDA0003173827430000021
And
Figure BDA0003173827430000022
adding them to obtain data W containing no cross termss′(t,w)；
Wherein the content of the first and second substances,
Figure BDA0003173827430000023
representing a chirp signal s1(t) the transformed signal;
Figure BDA0003173827430000024
representing a chirp signal s2(t) the transformed signal.
The invention provides a WVD time-frequency analysis cross item suppression system based on a generation countermeasure network, which comprises a storage, a processor and a computer program stored on the storage and capable of running on the processor, wherein when the program is executed by the processor, the steps of the method of any one of claims 1-8 are realized.
The invention provides a storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 8.
The invention has the beneficial effects that:
the WVD time-frequency analysis cross item suppression method based on the generation countermeasure network combines a deep learning network, and utilizes the end-to-end characteristic of the network and the convenience and diversity of cross item data preparation; the cross terms generated by the multi-component frequency modulation signal in the WVD conversion are eliminated. After the network provided by the method completes training, the network can directly carry out preliminary elimination of cross terms on the measured data, which is quicker compared with the traditional method. Under the condition of high signal-to-noise ratio, the method can be continuously exerted only by changing the training set, and is more accurate than the cross item elimination used by the traditional method. Images containing various interferences and noises can be trained by using the pix2pix network, and the stability of the final result is greatly improved.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention may be realized and attained by the means of the instrumentalities and combinations particularly pointed out hereinafter.
Drawings
In order to make the object, technical scheme and beneficial effect of the invention more clear, the invention provides the following drawings for explanation:
fig. 1 is a schematic diagram of the plate-like distribution (component number 2) of cross terms in the conversion of chirp signals WVD.
Fig. 2 is a schematic diagram of the linear distribution (component number 2) of cross terms in the conversion of the chirp signal WVD.
FIG. 3 is a diagram of a more complex cross term distribution.
Fig. 4 is a diagram of a GAN network architecture.
Fig. 5 is a view showing the structure of a pix2pix network.
Fig. 6 is a diagram of a generator network architecture.
Fig. 7 is a diagram of a discriminator network.
FIG. 8 is a flow chart of data creation with cross terms.
FIG. 9 is a flow chart of data creation without cross terms.
FIG. 10 is a comparison of image data with and without cross terms.
Fig. 11 is a graph showing the training effect of the training set.
FIG. 12 is the results of the test set.
Detailed Description
The present invention is further described with reference to the following drawings and specific examples so that those skilled in the art can better understand the present invention and can practice the present invention, but the examples are not intended to limit the present invention.
As shown in fig. 1, in the WVD (wigner-ville distribution) conversion according to this embodiment, the conversion of the chirp signal WVD is performed according to the following steps:
the WVD of the analytic signal z (n) is defined as follows:
Figure BDA0003173827430000041
in the formula, product term
Figure BDA0003173827430000042
Symmetrical about t. Unlike linear time-frequency analysis, the signal appears twice in the calculation formula of WVD, so that WVD is quadratic and nonlinear, and may also be referred to as bilinear transformation.
WVD is non-linear, i.e. the sum of two signals is not equal to the sum of WVD of each signal, and the extra part is the cross interference term. If s (t) ═ s1(t)+s2(t), then:
WVDs(t,w)＝WVDs1(t,w)+WVDs2(t,w)+2Re{WVDs1,s2(t,w)}#(2)
wherein
Figure BDA0003173827430000043
The first two terms of equation (2) represent the self term, i.e., useful information, of the signal, and the third term is the term of the added cross interference. The formula (2) shows that the cross interference term is real-valued, and not only is the cross interference term sandwiched between self terms, but also the amplitude of the cross interference term is twice the amplitude of the self terms, and a useful signal is inevitably drowned when the interference term is excessive; in addition, cross interferenceThe term is in the form of oscillation, and a cross interference term appears between every two signals, for example, a signal contains N components, and then the cross interference term is generated
Figure BDA0003173827430000044
A cross interference term.
Cross term distribution in the WVD conversion of the plurality of linear frequency modulation signals is respectively drawn as follows:
as shown in fig. 1, fig. 1 is a schematic diagram of the plate-like distribution (component number 2) of the cross terms in the conversion of the chirp signal WVD, and it can be seen that the cross terms are plate-like distributed when the chirp rates of the chirp signals are different greatly.
As shown in fig. 2, fig. 2 is a schematic diagram of the linear distribution (component number 2) of the cross terms in the conversion of the chirp signal WVD, and it can be seen that the cross terms are linearly distributed when the chirp rate of the chirp signal is small.
As shown in fig. 3, fig. 3 is a more complex cross term distribution diagram, when the component of the chirp signal increases, the cross term becomes more complex, and it can be seen that the cross terms of slice and line are distributed in the image.
In this embodiment, in order to eliminate the cross terms in the WVD varying image, the image containing the cross terms is processed by using a countermeasure generation network (GAN), and the idea is simplified as follows: a Generative antagonistic GAN network (GAN) is constructed, which includes two major network structures, generator G and discriminator D: wherein, the generator g (generator): responsible for making data out according to the blank; arbiter d (discriminator): and the game process of the two networks can eliminate cross items in the WVD change image. The traditional method for eliminating the cross terms is not completely eliminated when the multi-component frequency modulation signals are confronted, because the intensity, the frequency modulation slope and the initial slope of the multi-component signals can have great influence on the form distribution of the cross terms, and the generation of the countermeasure network for eliminating the cross terms can eliminate the influence by utilizing the diversity and the convenience of simulation data, which is the greatest advantage of a neural network compared with the traditional image processing.
The embodiment adopts a pix2pix network to eliminate the cross terms in the WVD variation image, wherein the pix2pix network is a branch in the GAN network and essentially maps pixels to pixels. The concrete structure is as follows:
as shown in fig. 4, fig. 4 is a diagram of a GAN network structure, a random noise z is generated by a generator G to obtain a generated sample G (z), the generated sample G (z) and real data x are input into a discriminator D, and a discrimination result is obtained by the discriminator D. In one iteration, firstly, the parameters of the discriminator D are trained, random noise z is input into the generator G to generate a sample G (z), and is transmitted into the discriminator D together with real data x, and the learning goal of the discriminator is to output 0 to the generated sample G (x) and output 1 to the real data x. Training the parameters of the generator next after the parameters of the discriminator D are trained, generating a sample G (x), obtaining a discrimination result in the discriminator D, and adjusting the parameters according to the discrimination result by the generator G, so that the output result is 1, namely the real data, namely the discriminator is cheated. And when the discriminator cannot distinguish the generated sample from the real data, the network is fitted.
As shown in fig. 5, fig. 5 is a view of a pix2pix network structure, a generator G obtains a generated sample G (x) of real data x, a discriminator D discriminates the generated sample G (x) as false (fake) under the condition of the real data x, and the discriminator D discriminates the real sample as real (real) under the condition of the real data x. Wherein x is data containing cross terms, y is data containing no cross terms, and the two images pass through a discriminator to obtain a positive sample real. And X passes through a generator G to obtain an image G (X) with the cross terms eliminated, and then the image with the cross terms eliminated and the image with the cross terms are input into a discriminator D to be used as a negative sample fake.
As shown in fig. 6, fig. 6 is a network structure diagram of a generator, wherein the generator g (x) adopts U-Net, which is a very wide network structure applied in the field of image segmentation and can fully fuse features, the generator comprises an down-sampling coding unit and an up-sampling coding unit, and the generator is performed according to the following steps: firstly, an input picture is down-sampled to a low-dimensional picture through a sampling coding unit, and then the low-dimensional picture is up-sampled to the original resolution through an up-sampling coding unit; and the down sampling and the up sampling are spliced according to corresponding channels through a connecting unit, so that the detail information of the pixel level under different resolutions is reserved. Therefore, the effect of U-Net on detail improvement is very obvious, and information of different scales can be reserved.
As shown in fig. 7, fig. 7 is a diagram of a network structure of a discriminator, and the discriminator provided in this embodiment uses a PatchGAN, which outputs a prediction probability value for each region (patch) of an input image, which corresponds to a determination of whether an input is true or false, to a determination of whether an input region of a certain size is true or false. When the method is specifically realized, a 1024 x 1024 input full convolution small network is used, the probability that each pixel output is true after sigmoid output in the last layer is output, and then BCEloss is used for calculating to obtain the final loss. This has the advantage that because the input dimension is greatly reduced, the parameters are reduced, the operation speed is faster than that of directly inputting, and the graph with any size can be calculated.
This example produces simulation data by:
the image containing the cross terms is used as random noise and input into a generator, the image completely eliminating the cross terms is used as real data and input into a discriminator, and the generator and the discriminator are mutually confronted, so that the aim of eliminating the cross terms is fulfilled.
The data making process comprises the following steps:
FIG. 8 is a flow chart of data generation with cross terms, as shown in FIG. 8, for applying a chirp signal s, as shown in FIG. 81(t) and s2(t) adding to obtain s (t), and then carrying out WVD conversion on s (t) to obtain data W containing cross termss(t, w); s (t) represents the added signal; w is as(t, w) represents a signal after WVD conversion; the cross terms provided by the embodiment are nonlinear parts generated when the chirp signal is subjected to WVD change, and the main purpose of the embodiment is to eliminate the cross terms generated in the process.
As shown in FIG. 9, FIG. 9 is a data system without cross terms as shown in FIG. 9As a flow chart, the chirp signals s are respectively converted1(t) and s2(t) performing a WVD conversion to obtain
Figure BDA0003173827430000061
And
Figure BDA0003173827430000062
adding them to obtain data W containing no cross termss' (t, w); wherein the content of the first and second substances,
Figure BDA0003173827430000063
representing a chirp signal s1(t) the transformed signal;
Figure BDA0003173827430000064
representing a chirp signal s2(t) the transformed signal;
as shown in fig. 10, fig. 10 is a schematic diagram showing comparison between an image containing cross terms and image data not containing cross terms, where (a) is data containing cross terms, and (b) is data not containing cross terms, and simulated data includes cross term distributions in a sheet shape and a line shape. The linear frequency modulation signal comprises a plurality of components, wherein the number of the components is randomly given to be 2-4; the amplitude of each component signal is randomly given as a constant value, and the amplitude is within the range of 1-2; the initial slope is randomly set within the range of 1-200. Because the influence of the frequency modulation slope on the cross terms of the WVD conversion is large, when the frequency modulation slopes of the two signal components are relatively close, the cross terms are in linear distribution, and otherwise, the cross terms are in sheet distribution. Therefore, the slice-shaped data and the line-shaped data account for 50% of the simulation data.
In the WVD time-frequency analysis cross term suppression method based on generation of an antagonistic network provided in this embodiment, a result of prediction performed by the network is shown in fig. 11, where fig. 11 is a training effect graph of a training set, where (a) is data containing cross terms, (b) is a result obtained after training of the training set is completed, and (c) is a label image.
The WVD time-frequency analysis cross item suppression method based on the generated countermeasure network provided by this embodiment is quite excellent in performance on the test set, and the test result is shown in fig. 12, where (a) is the result containing cross item data, (b) is the result obtained by the test set, and (c) is the label image, and as can be seen from the test result, the cross items, whether in a sheet shape or a line shape, are eliminated well, and the visual effect is quite ideal.
The above-mentioned embodiments are merely preferred embodiments for fully illustrating the present invention, and the scope of the present invention is not limited thereto. The equivalent substitution or change made by the technical personnel in the technical field on the basis of the invention is all within the protection scope of the invention. The protection scope of the invention is subject to the claims.

Claims (10)

1. The WVD time-frequency analysis cross item suppression method based on the generation countermeasure network is characterized by comprising the following steps: the method comprises the following steps:
constructing a generative antagonistic GAN network, said GAN network comprising a generator G and an arbiter D: the generator G is used for acquiring sample data; the discriminator D is used for judging whether the input data is real data;
acquiring an image containing a cross item and an image for eliminating the cross item;
the generator and the discriminator are trained in a mutual game until the cross terms in the image containing the cross terms are eliminated.
2. The WVD time-frequency analysis cross term suppression method based on generation of countermeasure networks according to claim 1, wherein: the method for training the generator and the discriminator in a mutual game mode comprises the following specific steps:
training parameters of a discriminator D:
generating a sample G (z) by inputting the image containing the cross terms into a generator G;
inputting the sample G (z) and the image for eliminating the cross terms into a discriminator D for training to obtain parameters of the discriminator D;
outputting the discrimination result of the discriminator D;
parameter training of the generator:
adjusting the parameters of the generator according to the judgment result output by the judger D until the output result of the generator G is in a real state;
the parameters of generator G are retained.
3. The WVD time-frequency analysis cross term suppression method based on generation of countermeasure networks according to claim 1, wherein: the generative antagonistic GAN network employs a pix2pix network.
4. The WVD time-frequency analysis cross term suppression method based on generation of countermeasure networks according to claim 1, wherein: the generator G (x) adopts a U-Net network and comprises a down-sampling coding unit, an up-sampling coding unit and a connecting unit;
the U-Net network is carried out according to the following steps:
down-sampling an input image to a low-dimensional image through a sampling coding unit, and up-sampling the low-dimensional image to an original resolution through an up-sampling coding unit;
and splicing the down sampling and the up sampling according to corresponding channels through a connecting unit.
5. The WVD time-frequency analysis cross term suppression method based on generation of countermeasure networks according to claim 1, wherein: the discriminator adopts Markov discriminator PatchGAN.
6. The WVD time-frequency analysis cross term suppression method based on generation of countermeasure networks according to claim 5, wherein: the full convolution in the Markov discriminator PatchGAN adopts a full convolution small network, the probability that each pixel in the last layer outputs true after passing sigmoid is calculated by BCEloss to obtain the final loss.
7. The WVD time-frequency analysis cross term suppression method based on generation of countermeasure networks according to claim 1, wherein: the image containing the cross terms is manufactured according to the following steps:
by applying a linear frequency-modulated signal s1(t) and s2(t) adding to obtain s (t), and then performing WVD conversion on s (t)In other words, data W containing cross terms is obtaineds(t,w)。
8. The WVD time-frequency analysis cross term suppression method based on generation of countermeasure networks according to claim 1, wherein: the cross item eliminating image is manufactured according to the following steps:
separately applying a linear frequency-modulated signal s1(t) and s2(t) performing WVD conversion to obtain Ws1(t, W) and Ws2(t,w)，
Adding them to obtain data W containing no cross termss′(t,w)；
Wherein, ws1(t, w) represents a chirp signal s1(t) the transformed signal; w is as2(t, w) represents a chirp signal s2(t) the transformed signal.
9. The WVD time-frequency analysis cross item suppression system based on the generation countermeasure network comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, and is characterized in that: the program when executed by a processor implements the steps of the method of any one of claims 1 to 8.
10. A storage medium having a computer program stored thereon, characterized in that: the program when executed by a processor implements the steps of the method of any one of claims 1 to 8.
CN202110826554.7A 2021-07-21 2021-07-21 WVD time-frequency analysis cross item suppression method and system based on generation countermeasure network and storage medium Pending CN113572710A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110826554.7A CN113572710A (en) 2021-07-21 2021-07-21 WVD time-frequency analysis cross item suppression method and system based on generation countermeasure network and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110826554.7A CN113572710A (en) 2021-07-21 2021-07-21 WVD time-frequency analysis cross item suppression method and system based on generation countermeasure network and storage medium

Publications (1)

Publication Number Publication Date
CN113572710A true CN113572710A (en) 2021-10-29

Family

ID=78166015

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110826554.7A Pending CN113572710A (en) 2021-07-21 2021-07-21 WVD time-frequency analysis cross item suppression method and system based on generation countermeasure network and storage medium

Country Status (1)

Country Link
CN (1) CN113572710A (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109948693A (en) * 2019-03-18 2019-06-28 西安电子科技大学 Expand and generate confrontation network hyperspectral image classification method based on super-pixel sample
CN110223254A (en) * 2019-06-10 2019-09-10 大连民族大学 A kind of image de-noising method generating network based on confrontation
US20200394475A1 (en) * 2019-06-13 2020-12-17 Visa International Service Association Method, System, and Computer Program Product for Generating New Items Compatible with Given Items

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109948693A (en) * 2019-03-18 2019-06-28 西安电子科技大学 Expand and generate confrontation network hyperspectral image classification method based on super-pixel sample
CN110223254A (en) * 2019-06-10 2019-09-10 大连民族大学 A kind of image de-noising method generating network based on confrontation
US20200394475A1 (en) * 2019-06-13 2020-12-17 Visa International Service Association Method, System, and Computer Program Product for Generating New Items Compatible with Given Items

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
严秦梦颖等: "基于条件对抗生成时频分布的多分量信号瞬时频率估计", 信号处理, vol. 35, no. 12, pages 3 *
俸玉祥;逯锋兵;石晓萍;葛宏宇;: "多分量线性调频信号时-频分析的交叉项抑制", 数字通信世界, no. 05, 1 May 2020 (2020-05-01), pages 91 - 92 *
俸玉祥等: "多分量线性调频信号时-频分析的交叉项抑制", 技术分析, 1 May 2020 (2020-05-01), pages 91 - 92 *
牛萌 等: "基于S 一霍夫变换的LFM信号时频检测", 计算机与网络, no. 11, 12 June 2009 (2009-06-12), pages 40 - 41 *
牛萌;刘光斌;: "基于S-Hough变换的多分量线性调频信号时频检测", 弹箭与制导学报, no. 03, 15 August 2007 (2007-08-15), pages 266 - 268 *
牛萌;李钊;李建军;胡礼勇;王天宝;: "基于S-霍夫变换的LFM信号时频检测", 计算机与网络, no. 11, 12 June 2009 (2009-06-12) *

Similar Documents

Publication Publication Date Title
Di Martino et al. Benchmarking framework for SAR despeckling
Rostami et al. Image deblurring using derivative compressed sensing for optical imaging application
CN104751420A (en) Sparse representation and multi-objective optimization based blind restoration method
CN113673312B (en) Deep learning-based radar signal intra-pulse modulation identification method
CN112837232A (en) Underwater image enhancement and detail recovery method
CN115601621A (en) Strong scattering medium active single-pixel imaging method based on condition generation countermeasure network
Mattausch et al. Image-based PSF estimation for ultrasound training simulation
CN116309062A (en) Remote sensing image super-resolution reconstruction method
CN113298700B (en) High-resolution image reconstruction method in scattering scene
Schüßler et al. Deep learning based image enhancement for automotive radar trained with an advanced virtual sensor
Wang et al. Local conditional neural fields for versatile and generalizable large-scale reconstructions in computational imaging
Guan et al. Source separation and noise reduction in single-pixel imaging
CN113378472A (en) Mixed boundary electromagnetic backscattering imaging method based on generation countermeasure network
Fan et al. Interference mitigation for synthetic aperture radar using deep learning
CN113572710A (en) WVD time-frequency analysis cross item suppression method and system based on generation countermeasure network and storage medium
CN115880158A (en) Blind image super-resolution reconstruction method and system based on variational self-coding
CN115223033A (en) Synthetic aperture sonar image target classification method and system
CN112529080B (en) Image generation method based on spectrum feature discrimination
CN114998137A (en) Ground penetrating radar image clutter suppression method based on generation countermeasure network
Chilukuri et al. Analysing Of Image Quality Computation Models Through Convolutional Neural Network
CN104063864A (en) Image fuzziness assessment method based on quaternary phase congruency model
Zhu et al. Quantitative assessment mechanism transcending visual perceptual evaluation for image dehazing
Ekmekci et al. Quantifying Generative Model Uncertainty in Posterior Sampling Methods for Computational Imaging
Wang et al. Enhancing time‐frequency resolution via deep‐learning framework
Estrada et al. Multi-frame image fusion using a machine learning-based weight mask predictor for turbulence-induced image degradation

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