CN112668498A - Method, system, terminal and application for identifying individual intelligent increment of aerial radiation source - Google Patents
Method, system, terminal and application for identifying individual intelligent increment of aerial radiation source Download PDFInfo
- Publication number
- CN112668498A CN112668498A CN202011621625.1A CN202011621625A CN112668498A CN 112668498 A CN112668498 A CN 112668498A CN 202011621625 A CN202011621625 A CN 202011621625A CN 112668498 A CN112668498 A CN 112668498A
- Authority
- CN
- China
- Prior art keywords
- radiation source
- identification
- signal
- layer
- 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.)
- Granted
Links
- 230000005855 radiation Effects 0.000 title claims abstract description 95
- 238000000034 method Methods 0.000 title claims abstract description 75
- 230000006870 function Effects 0.000 claims abstract description 75
- 238000012549 training Methods 0.000 claims abstract description 58
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 26
- 230000009466 transformation Effects 0.000 claims abstract description 22
- 238000010586 diagram Methods 0.000 claims abstract description 17
- 230000004927 fusion Effects 0.000 claims abstract description 16
- 238000011176 pooling Methods 0.000 claims description 35
- 238000000605 extraction Methods 0.000 claims description 32
- 238000012545 processing Methods 0.000 claims description 17
- 238000000354 decomposition reaction Methods 0.000 claims description 16
- 210000002569 neuron Anatomy 0.000 claims description 16
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000011049 filling Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 5
- 238000009825 accumulation Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 4
- 125000004122 cyclic group Chemical group 0.000 claims description 4
- 238000004821 distillation Methods 0.000 claims description 4
- 238000011068 loading method Methods 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000001934 delay Effects 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000002156 mixing Methods 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims description 2
- 230000001419 dependent effect Effects 0.000 abstract description 7
- 238000013500 data storage Methods 0.000 abstract description 2
- 238000012544 monitoring process Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000008092 positive effect Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000007873 sieving Methods 0.000 description 1
Images
Abstract
The invention belongs to the technical field of individual identification of radiation sources, and discloses an intelligent incremental identification method, a system, a terminal and application of an air radiation source individual, which are used for respectively extracting four characteristics of a fuzzy function, a bispectrum transformation, a Hilbert-Huang transformation and a short-time Fourier transformation from a received air radiation source ADS-B (broadcast automatic dependent surveillance) signal, and performing linear characteristic fusion on the characteristics to obtain a new characteristic diagram; carrying out classification identification on the radiation source individuals of known types through a convolutional neural network to obtain a network model; and for untrained class data, training is performed in an incremental learning mode, and intelligent incremental identification of the aerial radiation source individuals is realized. The invention can have good identification accuracy under lower signal-to-noise ratio, has good identification capability under different channels, has low dependence on a single characteristic, solves the problem of batch arrival of training data, greatly shortens the time required by training and reduces the space required by data storage.
Description
Technical Field
The invention belongs to the technical field of individual identification of radiation sources, and particularly relates to an intelligent incremental identification method, system, terminal and application of an individual aerial radiation source.
Background
At present: the individual identification of the radiation source refers to the identification of a target individual by extracting one or more modulation characteristics shown by the received signal. With the increasing number of aerial target individuals, whether the aerial target radiation source works normally or not and the enemy and my attributes of the aerial target are abnormal need to be judged, and the judgment needs to be carried out on the premise of individual identification of the aerial radiation source, so that how to judge quickly and accurately is an important means for realizing national safety.
Radiation source individual identification technology based on fingerprint information has been started in the last century. SaK, Lang D, et al propose a method for extracting fingerprint information using constellation and inputting the fingerprint information into convolutional neural network for Identification, but the method has strict requirements for time synchronization information (SaK, Lang D, Wang C, et al. Song C, Xu J et al propose a non-stationary signal analysis method of empirical mode decomposition (empirical mode decomposition), but this method has a problem of modal aliasing (Song C, Xu J, Zhan Y.A method for specific experimental identification based on empirical mode decomposition [ C ]// IEEE International Conference Wireless communications. IEEE, 2010.). Gok G, Alp Y K et al propose a Method based on Variational Modal Decomposition (VMD) to identify different radiation source signals using envelope and instantaneous frequency of the received signal as a set of models, but this Method is complicated and not good for practical operation (Gok G, Alp Y K, Arikan O.A New Method for Specific Identification With resources on Real radiation Measurements [ J ]. IEEE Transactions on Information dynamics and Security, 2020, PP (99): 1-1.). Shieh C S et al propose statistics based on radiation source conventional parameters direction of arrival (DOA), Pulse Width (PW), Pulse Repetition Frequency (PRF) and Radar Frequency (RF) as a basis for classification identification, but this method has poor identification accuracy at low signal-to-noise ratios (Shieh C S, Lin C T.A vector neural network for estimation identification [ J ]. IEEE Transactions on Antennas and Propagation, 2002, 50(8): 1120-. Xudan, Yang wave et al propose a Kernel Principal Component Analysis (KPCA) predictive learning method to solve the problem of processing different data clusters with complex non-linear distributions, but the method is difficult to be applied to other communication scenarios under non-gaussian channels (Dan Xu, Bo Yang, Wenli Jiang, An improved SVDU-IKPCA algorithm for Specific implementation Identification [ C ]// International Conference on Information & automation, ieee, 2008.). Chenyue et al proposed communication radiation source individual identification based on IQ map features, but the features extracted by this method were not obvious and performed poorly at low signal-to-noise ratios (Chenyue, Leipike, LiXin, leaf bell, Meifan. communication radiation source individual identification based on IQ map features [ J/OL ]. Signal processing). Chenpeng et al propose a method of calculating inter-Signal distance, pulse envelope or instantaneous frequency using Frechet distance to achieve radiation source individual identification, which also has poor identification performance at low Signal-to-noise ratio (P.Chen, G.Li, K.xu and J.Wan., "Applying the front distance to the specific identifier," 2016IEEE 13th International Conference on Signal Processing (ICSP), Chengdu, 2016, pp.1027-1030.). D 'Agustino S proposes an algorithm for extracting the characteristics of signals by using short-time discrete Fourier transform (STDFT), and finally, classification and identification are carried out by using a clustering algorithm, but the method has large dependence on sliding window selection and lower identification accuracy of the clustering algorithm (D' Agustino S. specific estimation based on amplitude components [ C ]// IEEE International Signal & Image Processing applications. IEEE, 2015.).
The method solves the problem of individual identification of the radiation source to a certain extent, but the identification performance is poor in the environment with low signal to noise ratio, the generalization capability is insufficient in the environment with different channels, the individual identification method has strong dependence on the feature extraction method, and the identification accuracy of different feature extraction methods is greatly different. In addition, the conventional training strategy must require all data classes as a priori information and perform uniform training. However, in reality, data is not collected into a whole block but is streamed, and the traditional joint training mode is not suitable for the current environment.
Through the above analysis, the problems and defects of the prior art are as follows:
(1) the prior art has the problems that the identification performance is poor under the environment with low signal-to-noise ratio, the generalization capability is insufficient under the environment with different channels, the dependence of an individual identification method on a feature extraction method is strong, the identification accuracy of different feature extraction methods is greatly different, and the like.
(2) Conventional training strategies must require that all data classes be known in advance and uniformly trained. However, in reality, data is not collected into a whole block but is streamed, and the traditional joint training mode is not suitable for the current environment.
The difficulty in solving the above problems and defects is:
(1) due to the influence of the current complex electromagnetic environment and channels, the individual signals of common aerial radiation sources are difficult to achieve a high signal-to-noise ratio, the signal-to-noise ratio of the signals is low, the interference is strong, the signals are normal, and a feature extraction method which is strong in anti-interference capability and good in recognition capability on most channels is difficult to find;
(2) the training of streaming data needs to keep the past data, the storage of old data has a large demand for memory space, and if all the data are stored and retrained, not only the previous training results are wasted, but also a large amount of memory space is wasted, which is not favorable for the training speed improvement and the resource maximization utilization.
The significance of solving the problems and the defects is as follows:
the types of aerial radiation source individuals are more and more, signals which are staggered with each other are more and more dense, the effect of interference signals of enemy countries is stronger and stronger, the battlefield form is changeable instantly, and the whole battlefield form trend can be influenced by identifying the radiation source individuals more quickly and accurately. Therefore, how to accurately identify the aerial radiation source individuals in a complex electromagnetic environment with low signal-to-noise ratio and strong interference, and quickly update an individual identification system by using the previous training result and new data so as to improve the capability of electronic warfare becomes a key problem to be solved urgently in modern warfare.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides an air radiation source individual intelligent increment identification method, system, terminal and application.
The invention is realized in such a way that an intelligent increment identification method for an individual aerial radiation source comprises the following steps:
respectively extracting four characteristics of a fuzzy function, a bispectrum transformation, a Hilbert-Huang transformation and a short-time Fourier transformation from a received ADS-B (broadcast automatic dependent surveillance) signal of an aerial radiation source, and performing linear characteristic fusion on the characteristics to obtain a new characteristic diagram;
carrying out classification identification on the radiation source individuals of known types through a convolutional neural network to obtain a network model;
and for untrained class data, training in an incremental learning mode, namely, on the basis of an original training result, jointly training the network by using a small part of old data and new data, and modifying partial parameters and loss functions in the network to realize intelligent incremental identification of the aerial radiation source individuals.
Further, the four feature extraction methods are specifically implemented as follows:
(1) short-time fourier transform STFT: the short-time fourier transform expression is:
wherein the subscript n is distinct from the standard fourier transform, w (n-m) is a sequence of window functions, x (m) is an input sequence, different sequences of window functions will yield different fourier transform results; the short-time Fourier transform has two independent variables n and w, namely a discrete function related to time n and a continuous function related to angular frequency w;
(2) fuzzy functions are used to analyze and design various signals to identify different types of signals. Assuming a time delay difference of τ, a frequency shift of μ, and an input signal of X1(t)、X2(t), the ambiguity function is defined as:
(3) the bispectrum of a received signal r (n) is estimated by a nonparametric method, r (n) is firstly divided into gamma sections, each section comprises delta samples, and the third-order cycle cumulant of the received signal r (n) is written as:
wherein, χγ(τ1,τ2) Is the third order cyclic accumulation of each signal, tau1,τ2For different time delays, w (τ)1,τ2) Is a hexagonal window function, the bispectral estimate of the signal r (n) is expressed as:
(4) hilbert yellow transform, a method that combines empirical mode decomposition with hilbert transform. The basic principle of the empirical mode decomposition algorithm is as follows: first, find the upper envelope X of the original signal X (t)max(t) and the lower envelope Xmin(t), and averaging the upper and lower envelope lines:
secondly, the original signal X (t) and the average envelope m1(t) subtracting to obtain a residual signal d1(t); for the rest signal d1(t) repeating the above operations until the value is less than the screening threshold (SD), and obtaining the final proper first-order modal component c1(t), the first IMF, where the SD solution is as follows:
thirdly, for signals X (t) and c1(t) obtaining a first order residual amount r by subtracting1(t) adding r1(t) replacing the original signal X (t) to perform the above operation, repeating the operation n times to obtain an n-order mode function cn(t) and the amount of residual r that ultimately meets the criterian(t), the expression of the original signal x (t) after empirical mode decomposition is:
and finally, performing time-frequency processing on the signal subjected to the empirical mode decomposition processing by using Hilbert transform, and converting the signal into a Hilbert spectrogram.
Further, the implementation of the linear feature fusion of the features to obtain a new feature map is as follows: expanding the feature graph to the same size by adopting an interpolation zero-filling method, and then performing linear feature fusion on the four features to finally obtain a new feature graph; if the individual differences of the radiation sources are not obvious by a certain characteristic extraction method, the neural network judges which individual radiation source belongs to through other characteristics.
Further, the process of classifying and identifying the known classes of radiation source individuals through the convolutional neural network comprises the following steps:
firstly, feature extraction is carried out on an input feature map by a convolutional layer, the convolutional neural network generally processes picture data in the past, four modes of feature extraction are carried out on signals, linear feature fusion is carried out, each signal can obtain a four-channel data feature map, and the method is also suitable for the processing of the convolutional neural network. The convolution layer internally comprises a plurality of convolution kernels, each element forming each convolution kernel corresponds to a weight coefficient and a deviation value, each neuron in each convolution layer is connected with a plurality of neurons of an area close to the position in the previous layer, and the size of the area depends on the size of the convolution kernels. The calculation formula is as follows:
wherein b is a deviation amount, ω is a weight value, x and y denote convolution kernels for convolution processing of the entire characteristic diagram, and ZlAnd Zl+1The convolutional inputs and outputs representing the L +1 th layer, also called the signature, Ll+1Is Zl+1Assuming that the feature maps are equal in length and width; z (i, j) corresponds to the pixels of the feature map, K is the number of feature map channels, f is the convolution kernel size, s0Is the convolution step size, p is the fill size; the convolutional layer contains an excitation function to assist in expressing complex features, and a linear rectification function RELU is used to make neurons in the neural network have sparse activation, which is represented as follows:
secondly, after the feature extraction is carried out on the convolutional layer, the output features are transmitted to a pooling layer for feature selection and information filtering; adding pooling operation after the convolutional layer, wherein the step of selecting a pooling area by the pooling layer is the same as the step of scanning the characteristic diagram by the convolutional core, and the pooling size, the step length and the filling are controlled, and the expression is as follows:
in the right formula AkRepresenting the input signature, and other parameters are associated with the convolutional layer.
And finally, inputting the output of the pooling layer into a full-connection layer, and carrying out nonlinear combination on the extracted features to obtain output.
Further, training the untrained category data in an incremental learning mode; on the basis of an original training result, a small part of old data and new data are taken to train a network together, and partial weight values and loss functions in the network are modified, so that intelligent incremental identification of the aerial radiation source individuals is realized; the method specifically comprises the following steps:
firstly, obtaining network model parameter information of a known class sample;
secondly, constructing a new data set by the old sample set and the new unknown class sample; mixing a small part of old data and new data to form a new data set, and disordering and randomly selecting an original data set by adopting a random selection principle on the selection of an old sample;
and finally, loading an old network model, modifying a network output layer, changing the number of output nodes into the number of radiation source individuals actually trained, reducing the step length of each layer of convolution kernel of the convolution layer and reducing the learning rate when the training recognition rate is low, adding dropout or a regular term for correction when an overfitting condition occurs, and optimizing a loss function as follows:
wherein the first term is the classifier loss function, the second term is the distillation loss function, qiRepresent the nodes of different neurons and are represented by,
representing a classifier.
It is a further object of the invention to provide a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of:
respectively extracting four characteristics of a fuzzy function, a bispectrum transformation, a Hilbert-Huang transformation and a short-time Fourier transformation from a received ADS-B (broadcast automatic dependent surveillance) signal of an aerial radiation source, and performing linear characteristic fusion on the characteristics to obtain a new characteristic diagram;
carrying out classification identification on the radiation source individuals of known types through a convolutional neural network to obtain a network model;
and for untrained class data, training in an incremental learning mode, namely, on the basis of an original training result, jointly training the network by using a small part of old data and new data, and modifying partial parameters and loss functions in the network to realize intelligent incremental identification of the aerial radiation source individuals.
It is another object of the present invention to provide a computer-readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:
respectively extracting four characteristics of a fuzzy function, a bispectrum transformation, a Hilbert-Huang transformation and a short-time Fourier transformation from a received ADS-B (broadcast automatic dependent surveillance) signal of an aerial radiation source, and performing linear characteristic fusion on the characteristics to obtain a new characteristic diagram;
carrying out classification identification on the radiation source individuals of known types through a convolutional neural network to obtain a network model;
and for untrained class data, training in an incremental learning mode, namely, on the basis of an original training result, jointly training the network by using a small part of old data and new data, and modifying partial parameters and loss functions in the network to realize intelligent incremental identification of the aerial radiation source individuals.
The invention also aims to provide an information data processing terminal, which is used for realizing the intelligent increment identification method of the aerial radiation source individuals.
Another object of the present invention is to provide an intelligent incremental identification system for implementing the method for identifying an individual intelligent increment of an airborne radiation source, wherein the intelligent incremental identification system comprises:
the characteristic combination module is used for carrying out Hilbert-Huang transform, short-time Fourier transform, fuzzy function and bispectrum transform characteristic extraction on the received ADS-B (broadcast automatic correlation monitoring) signal and combining the Hilbert-Huang transform, the short-time Fourier transform, the fuzzy function and the bispectrum transform into a new characteristic;
and the radiation source individual identification module is used for realizing radiation source individual identification through a convolutional neural network.
And the increment training module is used for carrying out increment training on the network by adopting an increment learning method.
The invention also aims to provide a radiation source individual identification terminal, which is used for realizing the intelligent increment identification method.
By combining all the technical schemes, the invention has the advantages and positive effects that: in order to identify the individual radiation source, the invention extracts the characteristics of the signals in four modes and fuses the linear characteristics, each signal can obtain a data 'characteristic diagram' of four channels, the invention is suitable for the application scene of the convolutional neural network, and after the four characteristic extraction modes are fused, each diagram contains the information of all characteristic extractions, thereby avoiding the dependence on a single characteristic. According to the invention, pooling operation is added after the convolutional layer, redundant information is filtered, and the training scale is reduced. The step of selecting the pooling area by the pooling layer is the same as the step of scanning the characteristic diagram by the convolution kernel, and the pooling size, the step length and the filling are controlled. The invention solves the problem of identification of individual radiation sources, so that a normalized exponential function (softmax function) output classification label is arranged at the last layer of a full-connection layer to obtain information such as network model parameters and the like.
The convolutional layer used in the present invention is processed with a larger convolutional kernel. Since AlexNet mainly operates on image classification, the addition of a pooling layer after each layer is to reduce redundant information, while the feature extracted by one step of the present invention is valid information, if a pooling layer is added after each layer of convolutional layer, part of the information may be lost. Therefore, no pooling layer needs to be added after each convolutional layer, but a global pooling layer is added after all convolutional layers, because the network has sufficient mining of the radiation source information, and the purpose of adding the global pooling is only to reduce the data size and accelerate the network convergence speed. Experiments prove that the network structure arrangement is really beneficial to individual identification of the radiation source.
The method obtains the information such as network model parameters of the known class samples and the like so as to facilitate incremental learning; second, a new data set is constructed from the old sample set and the new unknown class sample. In order to prevent the network from forgetting too much information of an old sample and prevent the network from being over-fitted with new category data, the method takes a small part of old data and new data to form a new data set, and adopts a random selection principle on the selection of the old sample to disorder and randomly select an original data set; and finally, loading an old network model, modifying a network output layer, changing the number of output nodes into the number of radiation source individuals actually trained, reducing the step length of each layer of convolution kernel of the convolution layer and reducing the learning rate when the training recognition rate is low, adding dropout or a regular term for correction when an overfitting condition occurs, optimizing a loss function in order to prevent the catastrophic forgetting problem in incremental learning, and increasing the consequences caused by training errors by setting a T value larger than 1, which is equivalent to load training, so that the training accuracy is higher. Combining these two terms yields a new loss function that is much better than using only the classifier loss function.
The invention can effectively realize the classification of ADS-B (broadcast automatic correlation monitoring) signals, has good identification performance in the environment with lower signal-to-noise ratio, reduces the dependency on the feature extraction method, and greatly reduces the training time and the space required by storage by adopting the incremental learning method. The invention provides an increment intelligent identification method of an individual aerial radiation source, which has good identification accuracy under the condition of a low signal-to-noise ratio, still has good identification capability under different channels, has low dependence on a single characteristic, solves the problem of batch arrival of training data, greatly shortens the time required by training and reduces the space required by data storage.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained from the drawings without creative efforts.
Fig. 1 is a flowchart of an individual intelligent incremental identification method for an airborne radiation source according to an embodiment of the present invention.
FIG. 2 is a schematic structural diagram of an individual intelligent incremental identification system for airborne radiation sources according to an embodiment of the present invention;
in fig. 2: 1. a characteristic combination module; 2. a radiation source individual identification module; 3. and an incremental training module.
Fig. 3 is a recognition accuracy chart provided by the embodiment of the present invention, which is input to a convolutional neural network after feature recombination and trained by an incremental learning method.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Aiming at the problems in the prior art, the invention provides an air radiation source individual intelligent increment identification method, a system, a terminal and application thereof, and the invention is described in detail below with reference to the accompanying drawings.
As shown in fig. 1, the method for identifying an individual intelligent increment of an airborne radiation source provided by the invention comprises the following steps:
s101: performing Hilbert-Huang transform, short-time Fourier transform, fuzzy function and bispectrum transform feature extraction on a received ADS-B (broadcast automatic dependent surveillance) signal, and combining the Hilbert-Huang transform, the short-time Fourier transform, the fuzzy function and the bispectrum transform into a new feature;
s102: the individual identification of the radiation source is realized through a convolutional neural network;
s103: and performing incremental training on the network by adopting an incremental learning method.
Those skilled in the art can also implement the intelligent increment identification method provided by the present invention by using other steps, and the intelligent increment identification method provided by the present invention in fig. 1 is only one specific embodiment.
As shown in fig. 2, the intelligent incremental identification system provided by the present invention includes:
the characteristic combination module 1 is used for carrying out Hilbert-Huang transform, short-time Fourier transform, fuzzy function and bispectrum transform characteristic extraction on a received ADS-B (broadcast automatic correlation monitoring) signal and combining the Hilbert-Huang transform, the short-time Fourier transform, the fuzzy function and the bispectrum transform into a new characteristic;
and the radiation source individual identification module 2 is used for realizing radiation source individual identification through a convolutional neural network.
And the increment training module 3 is used for carrying out increment training on the network by adopting an increment learning method.
The technical solution of the present invention is further described with reference to the following specific embodiments.
The intelligent increment identification method provided by the invention comprises the following steps:
firstly, performing feature extraction of the four methods on a received ADS-B (broadcast automatic dependent surveillance) signal and performing linear feature fusion to obtain a new feature map;
(1) short Time Fourier Transform (STFT)
The short-time fourier transform expression is:
subscript n differs from the standard fourier transform, w (n-m) is the window function sequence, x (m) is the input sequence, different window function sequences will yield different fourier transform results; the short-time fourier transform has two arguments n and w, which are both discrete functions with respect to time n and continuous functions with respect to angular frequency w.
(2) Fuzzy function (ambiguity function)
The blur function can be used to analyze and design various signals and can also be used to identify different types of signals. Assuming a time delay difference of τ and a frequency shift of μ, the outputThe incoming signal is X1(t)、X2(t), the ambiguity function is defined as:
(3) double spectrum transformation (Bi-spectrum transform)
The bispectrum of the received signal r (n) is estimated by a nonparametric method. r (n) is first divided into Γ segments, each segment containing Δ samples. The third order cyclic accumulation of the received signal r (n) can be written as:
wherein, χγ(τ1,τ2) Is the third order cyclic accumulation of each signal, tau1,τ2For different time delays, w (τ)1,τ2) Is a hexagonal window function, the bispectral estimate of the signal r (n) is expressed as:
(4) Hilbert-Huang transform
The hilbert-yellow transform is a method that combines empirical mode decomposition with the hilbert transform. The basic principle of the empirical mode decomposition algorithm is as follows:
first, find the upper envelope X of the original signal X (t)max(t) and the lower envelope Xmin(t), and averaging the upper and lower envelope lines:
secondly, the original signal X (t) and the average envelope m1(t) subtracting to obtain a residual signal d1(t) of (d). For the rest signal d1(t) repeating the above operations until the value is less than the sieving threshold (SD), at which time the final suitable value is obtainedOf the first order modal component c1(t), the first IMF.
The SD method is as follows:
thirdly, for signals X (t) and c1(t) obtaining a first order residual amount r by subtracting1(t) adding r1(t) replacing the original signal X (t) to perform the above operation, repeating the operation n times to obtain an n-order mode function cn(t) and the amount of residual r that ultimately meets the criterian(t), the expression of the original signal x (t) after empirical mode decomposition is:
and finally, performing time-frequency processing on the signal subjected to the empirical mode decomposition by using Hilbert transform, and converting the signal into a Hilbert spectrogram.
After extraction, linear feature fusion is carried out on the four features to obtain a new feature map.
And secondly, inputting the features obtained in the first step into a convolutional neural network for recognition.
The convolutional neural network is defined as: the convolutional neural network comprises a convolutional layer, a pooling layer, a full-link layer and an output layer.
The convolutional layer has the function of extracting the characteristics of input data, and the convolutional neural network usually processes picture data and has very strong characterization and classification discrimination capability. In order to identify the individual radiation source, the invention performs feature extraction in four modes and linear feature fusion on the signals as described in the first step, so that each signal can obtain a data 'feature map' of four channels, which is very suitable for the application scene of the convolutional neural network. First, the convolution layer performs "feature extraction" (distinguished from the feature extraction of the first step, which refers to information acquisition and learning on input data) on an input feature map, and the convolution layer includes a plurality of convolution kernels therein, and each element constituting a convolution kernel corresponds to a weight coefficient and a bias vector (bias vector), and is similar to a neuron (neuron) of a feedforward neural network. Each neuron in the convolution layer is connected to a plurality of neurons in a closely located region in the previous layer, the size of the region being dependent on the size of the convolution kernel. The calculation formula is as follows:
wherein b is a deviation amount, ω is a weight value, x and y denote convolution kernels for convolution processing of the entire characteristic diagram, and ZlAnd Zl+1Represents the convolution input and output of the l +1 th layer, also called feature map. L isl+1Is Zl+1Assuming that the feature maps are equal in length and width. Z (i, j) corresponds to a pixel of the feature map, K is the number of feature map channels, f is the convolution kernel size (Kernal size), s0Is the convolution step size (stride) and p is the padding size (padding). The convolutional layer contains an excitation function to assist in expressing complex characteristics, and the invention uses a linear rectification function RELU to make neurons in a neural network have sparse activation, and the expression form is as follows:
secondly, after the feature extraction is carried out on the convolutional layer, the output features are transmitted to the pooling layer for feature selection and information filtering. Due to the characteristics of the signal and the feature extraction method, the feature map extracted in the first step is generally large in size, although convolution layer processing is performed, the size of the feature is still large, and a lot of redundant information exists, so that the training speed is not improved favorably. Therefore, the invention adds pooling operation after the convolution layer, filters redundant information and reduces training scale. The step of selecting the pooling area by the pooling layer is the same as the step of scanning the characteristic diagram by the convolution kernel, and the pooling size, the step length and the filling are controlled. The expression is as follows:
in the right formula AkRepresenting the input signature, and other parameters are associated with the convolutional layer.
And finally, inputting the output of the pooling layer into a full connection layer, and carrying out nonlinear combination on the extracted features to obtain the output. The invention solves the problem of identification of individual radiation sources, so that a normalized exponential function (softmax function) output classification label is arranged at the last layer of a full-connection layer to obtain information such as network model parameters and the like.
The invention mainly uses the convolutional neural network to classify the individual radiation sources, and because of the strong classification capability of the AlexNet network on the image information, the invention uses the improved AlexNet network to classify and identify the individual radiation sources. Because the size of the data to be processed is large, the convolutional layer used in the present invention is processed by using a large convolutional kernel. Since AlexNet mainly operates on image classification, the addition of a pooling layer after each layer is to reduce redundant information, while the features extracted in the first step of the present invention are valid information, if a pooling layer is added after each layer of convolutional layer, part of the information may be lost. Therefore, no pooling layer needs to be added after each convolutional layer, but a global pooling layer is added after all convolutional layers, because the network has sufficient mining of the radiation source information, and the purpose of adding the global pooling is only to reduce the data size and accelerate the network convergence speed. Experiments prove that the network structure arrangement is really beneficial to individual identification of the radiation source.
The network provided by the invention comprises 6 million parameters and 65000 neurons, five convolutional layers and three full-connection layer networks, and the final output layer is a softmax layer. A first layer: convolution layer 1, the number of convolution kernels is 96, the size of the convolution kernel is 11 × 11, stride is 4, and extension padding is 0; a second layer: the convolutional layer 2 is input as a feature map of the previous layer of convolution, the number of convolutions is 256, and the size of convolution kernel is: 5, padding ═ 2, stride ═ 1,; and a third layer: the convolution layer 3 has the input of the output of the second layer, the number of convolution kernels is 384, the size of the convolution kernels is 3 x 3, and the padding is 1; a fourth layer: the convolutional layer 4 has the output of the third layer as input, the number of convolutional kernels is 384, the size of the convolutional kernels is 3 × 3, and padding is 1. And a fifth layer: the convolutional layer 5 receives the output of the fourth layer, has 256 convolutional kernels, and has a convolutional kernel size of 3 × 3 and padding of 1. Then directly carrying out maximum pooling, wherein the size of the pooling is 4 x 4, and stride is 2; the 6 th, 7 th and 8 th layers are all connected layers, the number of the neurons in each layer is 4096, and finally softmax is output. The full connectivity layer uses the RELU activation function and Dropout operation.
Thirdly, the incremental learning step is as follows: firstly, as described in the second step, information such as network model parameters of known class samples is obtained so as to facilitate incremental learning; second, a new data set is constructed from the old sample set and the new unknown class sample. In order to prevent the network from forgetting too much information of an old sample and prevent the network from being over-fitted with new category data, the method takes a small part of old data and new data to form a new data set, and adopts a random selection principle on the selection of the old sample to disorder and randomly select an original data set; and finally, loading an old network model, modifying a network output layer, changing the number of output nodes into the number of radiation source individuals actually trained, reducing the step length of each layer of convolution kernel of the convolution layer and reducing the learning rate when the training recognition rate is low, adding a dropout or a regular term for correction when an overfitting condition occurs, and optimizing a loss function as follows in order to prevent a catastrophic forgetting problem in incremental learning:
wherein the first term is the classifier loss function, the second term is the distillation loss function, qiRepresent the nodes of different neurons and are represented by,
representing a classifier. While
Through observation, the distillation coefficient T is added on the basis of the softmax function, and the T value larger than 1 is set, so that the result caused by training errors is aggravated, namely, the training is carried out under the condition of 'weight bearing', and the training accuracy is higher. Combining these two terms yields a new loss function that is much better than using only the classifier loss function. The invention has the advantages and positive effects that: the disclosed method for identifying the individual intelligent increment of the aerial radiation source can effectively realize the classification of ADS-B (broadcast automatic correlation monitoring) signals, has good identification performance in the environment with lower signal-to-noise ratio, reduces the dependency on a feature extraction method, and greatly reduces the training time and the space required by storage by adopting an increment learning method.
The invention verifies the method for identifying the aerial target radiation source through simulation experiments, simulation signals comprise the fingerprint information such as frequency deviation, phase distortion, harmonic distortion and the like mentioned in the foregoing, the simulation signals use ADS-B (broadcast automatic correlation monitoring) signals of S-Mode transponder Extended messages (1090ES, 1090MHz Mode S Extended Squitter), the working frequency is 1090MHz, the data rate is 1Mbps, and the modulation Mode is Pulse Position Modulation (PPM) and binary amplitude keying signals (2 ASK). The period of the signal is 120 mus, the leading pulse is 8 mus, the duration is 4 pulses of 0.5 mus, and the starting time is 0.1 mus, 3.5 mus and 4.5 mus respectively. The information pulse takes 112 mus, transmits 112 bits of data, one bit of data represents a message containing the position, altitude, speed, course, identification number and other information of the airplane, and the 01 and 10 binary data used after PPM modulation represents each message. The method includes the steps of intercepting data with the length of 5 mu s of ADSB signal leading pulse, sampling by using the frequency of 600MHz, setting the range of signal to noise ratio to be-5 dB, setting 3 different individuals, and generating 5 different signals for each individual, wherein the modulation frequency of target 1 phase noise is set to be 4MHz, 6MHz, 7MHz, 10MHz and 15MHz respectively, the phase modulation coefficients are set to be 0.16, 0.27, 0.32, 0.15 and 0.25 respectively, and the first harmonic component to fifth harmonic component are set to be 1, 0.5, 0.3, 0.2 and 0.1 respectively. The target 2 phase noise modulation frequencies are set to 2MHz, 5MHz, 9MHz, 11MHz, 13MHz, the phase modulation coefficients are set to 0.21, 0.32, 0.15, 0.24, 0.28, and the first to fifth harmonic components are set to 1, 0.8, 0.6, 0.4, 0.2, respectively. The target 3-phase noise modulation frequencies are set to 3MHz, 5MHz, 6MHz, 8MHz, 12MHz, the phase modulation coefficients are set to 0.34, 0.3, 0.23, 0.21, 0.26, and the first to fifth harmonic components are set to 1, 0.1, 0.08, 0.05, 0.03, respectively.
Fig. 3 shows the accuracy of the individual identification of the radiation source by using the four feature extraction methods in different combination methods, and it can be seen that the identification accuracy reaches 95% when the signal-to-noise ratio is greater than 3dB by using the method of combining the four features.
It should be noted that the embodiments of the present invention can be realized by hardware, software, or a combination of software and hardware. The hardware portion may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or specially designed hardware. Those skilled in the art will appreciate that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, such code being provided on a carrier medium such as a disk, CD-or DVD-ROM, programmable memory such as read only memory (firmware), or a data carrier such as an optical or electronic signal carrier, for example. The apparatus and its modules of the present invention may be implemented by hardware circuits such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of hardware circuits and software, e.g., firmware.
The above description is only for the purpose of illustrating the present invention and the appended claims are not to be construed as limiting the scope of the invention, which is intended to cover all modifications, equivalents and improvements that are within the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. An intelligent increment identification method for an individual airborne radiation source is characterized by comprising the following steps:
respectively extracting four characteristics of a fuzzy function, a bispectrum transformation, a Hilbert-Huang transformation and a short-time Fourier transformation from a received ADS-B signal of an aerial radiation source, and performing linear characteristic fusion on the characteristics to obtain a new characteristic diagram;
carrying out classification identification on the radiation source individuals of known types through a convolutional neural network to obtain a network model;
and for untrained class data, training in an incremental learning mode, namely, on the basis of an original training result, jointly training the network by using a small part of old data and new data, and modifying partial parameters and loss functions in the network to realize intelligent incremental identification of the aerial radiation source individuals.
2. The method for intelligently identifying individual increments of airborne radiation sources according to claim 1, wherein the four feature extraction methods are embodied as:
(1) short-time fourier transform STFT: the short-time fourier transform expression is:
wherein the subscript n is distinct from the standard fourier transform, w (n-m) is a sequence of window functions, x (m) is an input sequence, different sequences of window functions will yield different fourier transform results; the short-time Fourier transform has two independent variables n and w, namely a discrete function related to time n and a continuous function related to angular frequency w;
(2) fuzzy function for analyzing and designing various signals to identify different types of signals, with time delay difference of tau, frequency shift of mu, input signal of X1(t)、X2(t), the ambiguity function is defined as:
(3) the bispectrum of a received signal r (n) is estimated by a nonparametric method, r (n) is firstly divided into gamma sections, each section comprises delta samples, and the third-order cycle cumulant of the received signal r (n) is written as:
wherein, χγ(τ1,τ2) Is the third order cyclic accumulation of each signal, tau1,τ2For different time delays, w (τ)1,τ2) Is a hexagonal window function, the bispectral estimate of the signal r (n) is expressed as:
(4) Hilbert-Huang transform, a method combining empirical mode decomposition with Hilbert transform, the fundamental principle of the empirical mode decomposition algorithm is as follows: first, find the upper envelope X of the original signal X (t)max(t) and the lower envelope Xmin(t), and averaging the upper and lower envelope lines:
secondly, the original signal X (t) and the average envelope m1(t) subtracting to obtain a residual signal d1(t); for the rest signal d1(t) repeating the above operations until the value is less than the screening threshold (SD), and obtaining the final proper first-order modal component c1(t), the first IMF, where the SD solution is as follows:
thirdly, for signals X (t) and c1(t) obtaining a first order residual amount r by subtracting1(t) adding r1(t) replacing the original signal X (t) to perform the above operation, repeating the operation n times to obtain an n-order mode function cn(t) and the amount of residual r that ultimately meets the criterian(t), the expression of the original signal x (t) after empirical mode decomposition is:
and finally, performing time-frequency processing on the signal subjected to the empirical mode decomposition by using Hilbert transform, and converting the signal into a Hilbert spectrogram.
3. The method for identifying individual intelligent increments of airborne radiation sources according to claim 1, wherein the implementation of the linear feature fusion of the features to obtain a new feature map is as follows: expanding the feature graph to the same size by adopting an interpolation zero-filling method, and then performing linear feature fusion on the four features to finally obtain a new feature graph; if the individual differences of the radiation sources are not obvious by a certain characteristic extraction method, the neural network judges which individual radiation source belongs to through other characteristics.
4. The method for intelligently identifying the individual airborne radiation sources according to claim 1, wherein the process of classifying and identifying the known classes of individual radiation sources through the convolutional neural network comprises:
firstly, feature extraction is carried out on an input feature map by a convolutional layer, the convolutional neural network generally processes picture data in the past, four modes of feature extraction are carried out on signals, linear feature fusion is carried out, each signal can obtain a four-channel data feature map, and the method is also suitable for the processing of the convolutional neural network. The convolution layer internally comprises a plurality of convolution kernels, each element forming each convolution kernel corresponds to a weight coefficient and a deviation value, each neuron in each convolution layer is connected with a plurality of neurons of an area close to the position in the previous layer, the size of the area depends on the size of the convolution kernels, and the calculation formula is as follows:
wherein b is a deviation amount, ω is a weight value, x and y denote convolution kernels for convolution processing of the entire characteristic diagram, and ZlAnd Zl+1The convolutional inputs and outputs representing the L +1 th layer, also called the signature, Ll+1Is Zl+1Assuming that the feature maps are equal in length and width; z (i, j) corresponds to the pixels of the feature map, K is the number of feature map channels, f is the convolution kernel size, s0Is the convolution step size, p is the fill size; the convolutional layer contains an excitation function to assist in expressing complex features, and a linear rectification function RELU is used to make neurons in the neural network have sparse activation, which is represented as follows:
secondly, after the feature extraction is carried out on the convolutional layer, the output features are transmitted to a pooling layer for feature selection and information filtering; adding pooling operation after the convolutional layer, wherein the step of selecting a pooling area by the pooling layer is the same as the step of scanning the characteristic diagram by the convolutional core, and the pooling size, the step length and the filling are controlled, and the expression is as follows:
in the right formula AkRepresenting the input signature, and other parameters are associated with the convolutional layer.
And finally, inputting the output of the pooling layer into a full-connection layer, and carrying out nonlinear combination on the extracted features to obtain output.
5. The method for intelligently incrementally recognizing airborne radiation source individuals as claimed in claim 1, wherein the untrained category data is trained in an incremental learning manner; on the basis of an original training result, a small part of old data and new data are taken to train a network together, and partial weight values and loss functions in the network are modified, so that intelligent incremental identification of the aerial radiation source individuals is realized; the method specifically comprises the following steps:
firstly, obtaining network model parameter information of a known class sample;
secondly, constructing a new data set by the old sample set and the new unknown class sample; mixing a small part of old data and new data to form a new data set, and disordering and randomly selecting an original data set by adopting a random selection principle on the selection of an old sample;
and finally, loading an old network model, modifying a network output layer, changing the number of output nodes into the number of radiation source individuals actually trained, reducing the step length of each layer of convolution kernel of the convolution layer and reducing the learning rate when the training recognition rate is low, adding dropout or a regular term for correction when an overfitting condition occurs, and optimizing a loss function as follows:
wherein the first term is the classifier loss function, the second term is the distillation loss function, qiRepresent the nodes of different neurons and are represented by,
representing a classifier.
6. A computer device, characterized in that the computer device comprises a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to carry out the steps of:
respectively extracting four characteristics of a fuzzy function, a bispectrum transformation, a Hilbert-Huang transformation and a short-time Fourier transformation from a received ADS-B signal of an aerial radiation source, and performing linear characteristic fusion on the characteristics to obtain a new characteristic diagram;
carrying out classification identification on the radiation source individuals of known types through a convolutional neural network to obtain a network model;
and for untrained class data, training in an incremental learning mode, namely, on the basis of an original training result, jointly training the network by using a small part of old data and new data, and modifying partial parameters and loss functions in the network to realize intelligent incremental identification of the aerial radiation source individuals.
7. A computer-readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:
respectively extracting four characteristics of a fuzzy function, a bispectrum transformation, a Hilbert-Huang transformation and a short-time Fourier transformation from a received ADS-B signal of an aerial radiation source, and performing linear characteristic fusion on the characteristics to obtain a new characteristic diagram;
carrying out classification identification on the radiation source individuals of known types through a convolutional neural network to obtain a network model;
and for untrained class data, training in an incremental learning mode, namely, on the basis of an original training result, jointly training the network by using a small part of old data and new data, and modifying partial parameters and loss functions in the network to realize intelligent incremental identification of the aerial radiation source individuals.
8. An information data processing terminal, characterized in that the information data processing terminal is used for realizing the intelligent increment identification method for the aerial radiation source individuals as claimed in any one of claims 1 to 5.
9. An intelligent incremental recognition system for implementing the intelligent incremental recognition method according to any one of claims 1 to 5, wherein the intelligent incremental recognition system for the individual airborne radiation sources comprises:
the characteristic combination module is used for performing Hilbert-Huang transform, short-time Fourier transform, fuzzy function and bispectrum transform characteristic extraction on the received ADS-B signal and combining the characteristic extraction into a new characteristic;
the radiation source individual identification module is used for realizing radiation source individual identification through a convolutional neural network;
and the increment training module is used for carrying out increment training on the network by adopting an increment learning method.
10. A radiation source individual identification terminal is characterized in that the radiation source individual identification terminal is used for realizing the intelligent increment identification method of the aerial radiation source individual according to any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011621625.1A CN112668498B (en) | 2020-12-30 | 2020-12-30 | Method, system, terminal and application for identifying individual intelligent increment of aerial radiation source |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011621625.1A CN112668498B (en) | 2020-12-30 | 2020-12-30 | Method, system, terminal and application for identifying individual intelligent increment of aerial radiation source |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112668498A true CN112668498A (en) | 2021-04-16 |
| CN112668498B CN112668498B (en) | 2024-02-06 |
Family
ID=75412118
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011621625.1A Active CN112668498B (en) | 2020-12-30 | 2020-12-30 | Method, system, terminal and application for identifying individual intelligent increment of aerial radiation source |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112668498B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113435246A (en) * | 2021-05-18 | 2021-09-24 | 西安电子科技大学 | Radiation source individual intelligent identification method, system and terminal |
| CN113435245A (en) * | 2021-05-18 | 2021-09-24 | 西安电子科技大学 | Method, system and application for identifying individual airborne radiation source |
| CN113923088A (en) * | 2021-10-21 | 2022-01-11 | 天津光电通信技术有限公司 | Automatic 5G signal digital modulation mode identification method based on HLNN |
| CN114664290A (en) * | 2022-05-17 | 2022-06-24 | 深圳比特微电子科技有限公司 | Sound event detection method and device and readable storage medium |
| CN114925748A (en) * | 2022-04-20 | 2022-08-19 | 北京市商汤科技开发有限公司 | Model training and modal information prediction method, related device, equipment and medium |
| CN116559809A (en) * | 2023-03-28 | 2023-08-08 | 南京桂瑞得信息科技有限公司 | Radar radiation source individual identification method based on multi-source fusion network |
| CN117544963A (en) * | 2024-01-04 | 2024-02-09 | 四川大学 | Method and equipment for identifying radiation source of cross-mode communication signal based on FTGan-Yolo |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108090412A (en) * | 2017-11-17 | 2018-05-29 | 西北工业大学 | A kind of radar emission source category recognition methods based on deep learning |
| CN109492765A (en) * | 2018-11-01 | 2019-03-19 | 浙江工业大学 | A kind of image Increment Learning Algorithm based on migration models |
| GB201906560D0 (en) * | 2018-08-24 | 2019-06-26 | Petrochina Co Ltd | Method and apparatus for automatically extracting image features of electrical imaging well logging |
| KR20190110939A (en) * | 2018-03-21 | 2019-10-01 | 한국과학기술원 | Environment sound recognition method based on convolutional neural networks, and system thereof |
| CN110472545A (en) * | 2019-08-06 | 2019-11-19 | 中北大学 | The classification method of the power components image of taking photo by plane of knowledge based transfer learning |
| US20200302230A1 (en) * | 2019-03-21 | 2020-09-24 | International Business Machines Corporation | Method of incremental learning for object detection |
| CN112087774A (en) * | 2020-09-14 | 2020-12-15 | 桂林电子科技大学 | Communication radiation source individual identification method based on residual error neural network |
-
2020
- 2020-12-30 CN CN202011621625.1A patent/CN112668498B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108090412A (en) * | 2017-11-17 | 2018-05-29 | 西北工业大学 | A kind of radar emission source category recognition methods based on deep learning |
| KR20190110939A (en) * | 2018-03-21 | 2019-10-01 | 한국과학기술원 | Environment sound recognition method based on convolutional neural networks, and system thereof |
| GB201906560D0 (en) * | 2018-08-24 | 2019-06-26 | Petrochina Co Ltd | Method and apparatus for automatically extracting image features of electrical imaging well logging |
| CN109492765A (en) * | 2018-11-01 | 2019-03-19 | 浙江工业大学 | A kind of image Increment Learning Algorithm based on migration models |
| US20200302230A1 (en) * | 2019-03-21 | 2020-09-24 | International Business Machines Corporation | Method of incremental learning for object detection |
| CN110472545A (en) * | 2019-08-06 | 2019-11-19 | 中北大学 | The classification method of the power components image of taking photo by plane of knowledge based transfer learning |
| CN112087774A (en) * | 2020-09-14 | 2020-12-15 | 桂林电子科技大学 | Communication radiation source individual identification method based on residual error neural network |
Non-Patent Citations (2)
| Title |
|---|
| 何丽;韩克平;朱泓西;刘颖;: "双分支迭代的深度增量图像分类方法", 模式识别与人工智能, no. 02 * |
| 高欣宇;张文博;姬红兵;欧阳成;: "新型雷达辐射源识别", 中国图象图形学报, no. 06 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113435246A (en) * | 2021-05-18 | 2021-09-24 | 西安电子科技大学 | Radiation source individual intelligent identification method, system and terminal |
| CN113435245A (en) * | 2021-05-18 | 2021-09-24 | 西安电子科技大学 | Method, system and application for identifying individual airborne radiation source |
| CN113435245B (en) * | 2021-05-18 | 2023-06-30 | 西安电子科技大学 | Method, system and application for identifying individual aerial radiation source |
| CN113435246B (en) * | 2021-05-18 | 2024-04-05 | 西安电子科技大学 | Intelligent radiation source individual identification method, system and terminal |
| CN113923088A (en) * | 2021-10-21 | 2022-01-11 | 天津光电通信技术有限公司 | Automatic 5G signal digital modulation mode identification method based on HLNN |
| CN113923088B (en) * | 2021-10-21 | 2023-08-08 | 天津光电通信技术有限公司 | HLNN-based automatic 5G signal digital modulation mode identification method |
| CN114925748A (en) * | 2022-04-20 | 2022-08-19 | 北京市商汤科技开发有限公司 | Model training and modal information prediction method, related device, equipment and medium |
| CN114664290A (en) * | 2022-05-17 | 2022-06-24 | 深圳比特微电子科技有限公司 | Sound event detection method and device and readable storage medium |
| CN114664290B (en) * | 2022-05-17 | 2022-08-19 | 深圳比特微电子科技有限公司 | Sound event detection method and device and readable storage medium |
| CN116559809A (en) * | 2023-03-28 | 2023-08-08 | 南京桂瑞得信息科技有限公司 | Radar radiation source individual identification method based on multi-source fusion network |
| CN117544963A (en) * | 2024-01-04 | 2024-02-09 | 四川大学 | Method and equipment for identifying radiation source of cross-mode communication signal based on FTGan-Yolo |
| CN117544963B (en) * | 2024-01-04 | 2024-03-26 | 四川大学 | Method and equipment for identifying radiation source of cross-mode communication signal based on FTGan-Yolo |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112668498B (en) | 2024-02-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112668498A (en) | Method, system, terminal and application for identifying individual intelligent increment of aerial radiation source | |
| CN110865357B (en) | Laser radar echo signal noise reduction method based on parameter optimization VMD | |
| CN106685478B (en) | Frequency hopping signal parameter estimation method based on signal time-frequency image information extraction | |
| Ozturk et al. | RF-based low-SNR classification of UAVs using convolutional neural networks | |
| CN112566174B (en) | Abnormal I/Q signal identification method and system based on deep learning | |
| CN102254319A (en) | Method for carrying out change detection on multi-level segmented remote sensing image | |
| CN112560803A (en) | Radar signal modulation identification method based on time-frequency analysis and machine learning | |
| Zhou et al. | Specific emitter identification via bispectrum-radon transform and hybrid deep model | |
| CN111222442A (en) | Electromagnetic signal classification method and device | |
| CN110619264A (en) | UNet + + based microseism effective signal identification method and device | |
| Nuhoglu et al. | Image segmentation for radar signal deinterleaving using deep learning | |
| Du et al. | DNCNet: Deep radar signal denoising and recognition | |
| CN111046697A (en) | Adaptive modulation signal identification method based on fuzzy logic system | |
| CN111951611A (en) | ADS-B weak signal detection device and method based on multi-feature fusion | |
| Zhang | Optimization performance analysis of 1090ES ADS-B signal separation algorithm based on PCA and ICA | |
| Guven et al. | Classifying LPI radar waveforms with time-frequency transformations using multi-stage CNN system | |
| KR101427149B1 (en) | A method of recognizing complex PRI modulation type of radar signals based on 2-layer SVM using exponential moving average | |
| Liu et al. | Incremental learning based radio frequency fingerprint identification using intelligent representation | |
| CN114710215B (en) | Method for fast blind detection of frequency hopping signal | |
| CN115600101A (en) | Unmanned aerial vehicle signal intelligent detection method and device based on priori knowledge | |
| CN115809426A (en) | Radiation source individual identification method and system | |
| CN114841195A (en) | Avionics space signal modeling method and system | |
| Joseph et al. | FlightSense: A spoofer detection and aircraft identification system using raw ADS-B data | |
| CN116827486B (en) | Blind detection system and method for short-wave communication signals | |
| Govalkar et al. | Siamese Network based Pulse and Signal Attribute 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 |