CN113536897B - Aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis - Google Patents
Aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis Download PDFInfo
- Publication number
- CN113536897B CN113536897B CN202110597435.9A CN202110597435A CN113536897B CN 113536897 B CN113536897 B CN 113536897B CN 202110597435 A CN202110597435 A CN 202110597435A CN 113536897 B CN113536897 B CN 113536897B
- Authority
- CN
- China
- Prior art keywords
- time
- image
- gray
- analysis
- signal
- 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.)
- Active
Links
- 238000010586 diagram Methods 0.000 title claims abstract description 80
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000004458 analytical method Methods 0.000 title claims abstract description 62
- 239000013598 vector Substances 0.000 claims description 57
- 238000003708 edge detection Methods 0.000 claims description 34
- 238000012545 processing Methods 0.000 claims description 16
- 238000009826 distribution Methods 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000000877 morphologic effect Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 8
- 230000000694 effects Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000005530 etching Methods 0.000 claims description 6
- 238000005096 rolling process Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 4
- 230000014509 gene expression Effects 0.000 claims description 4
- 238000002360 preparation method Methods 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 abstract description 6
- 238000000926 separation method Methods 0.000 abstract description 2
- 238000013528 artificial neural network Methods 0.000 description 11
- 238000005516 engineering process Methods 0.000 description 9
- 238000013527 convolutional neural network Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 230000010363 phase shift Effects 0.000 description 4
- 230000005855 radiation Effects 0.000 description 4
- 238000003672 processing method Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 101100001674 Emericella variicolor andI gene Proteins 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000003628 erosive effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 101100272279 Beauveria bassiana Beas gene Proteins 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000001514 detection method Methods 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
- 238000013139 quantization Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
Abstract
The invention discloses an aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis, which can automatically identify each different modulation type from signals aliased by nine different modulation modes including digital and analog modulation modes. The method is based on a parameter fit and a small range variance. The method has the advantages of low cost, high speed and high recognition accuracy, is applied to recognition and separation of baseband aliasing signals in the field of wireless communication/radar/electronic countermeasure, and can realize low cost and quick recognition of the current electromagnetic environment.
Description
Technical Field
The invention relates to an aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis, belonging to the field of baseband signal processing in wireless communication/radar/electronic countermeasure.
Background
With the continuous progress in the field of electronic countermeasure in modern military war, the intensity of electromagnetic signals in electronic countermeasure is increasingly enlarged, and various novel radars are continuously put into use and are dominant, so that the modulation mode adopted by radar radiation source signals is more and more complex, such as frequency modulation, phase modulation, amplitude modulation and the like. The classification and identification process of the radar radiation source signals plays an irreplaceable role in the field of electronic countermeasure, and with the more flexible and changeable radar radiation source signals, the identification of the radiation source signals based on the traditional conventional parameters gradually cannot meet the increasingly complex electromagnetic signal environment.
The method comprises preprocessing a received signal to obtain a complex baseband signal containing in-phase components and quadrature components, taking the complex baseband signal as a data set of an input convolutional neural network model, adjusting the model structure and super parameters such as a convolution kernel, a step size, a feature map, an activation function and the like through multiple training, and carrying out feature extraction and identification on the communication signal by using the trained model; the identification and classification of seven digital communication signal types of 2FSK, 4FSK, BPSK, 8PSK, QPSK, QAM, 16 and QAM64 are realized.
The literature Yi Yunqing, lv Lequn, lu Yuanyuan, etc. the complex signal pattern modulation recognition technology based on neural network [ J ]. Electronic information countermeasure technology 2020,35 (6): 16-21 (reference 2) proposes a complex signal recognition technology based on BP neural network. Through analyzing and extracting the signal characteristics with high discrimination, three layers of BP neural networks are designed, the recognition performance of various complex communication modulation signals is studied in an important way, and a recognition algorithm based on the BP neural networks is realized.
The literature' Gui Xiangsheng, hong Juting, dahua, and the like, a signal modulation mode identification method [ J ] based on a convolutional neural network, a modern computer, 2019 (10): 18-22,26 "(reference 3), proposes a modulation mode automatic identification method based on a Convolutional Neural Network (CNN). The convolutional neural network has feature extraction and self-learning capabilities, can grasp the fine features of signals from the angle of image processing, does not need human intervention or data statistics, and achieves the effect of classification and identification. The constellation diagram of the digital modulation signal contains important parameter information of the signal, which is a key feature for distinguishing different modulation modes.
Document Xiong Kunlai, luo Jingqing, wu Shilong LPI radar signal modulation identification based on time-frequency images and neural networks [ J ] rocket and guidance bulletin 2011,31 (5): 230-233 "(reference 4) proposes a new method of LPI radar signal identification based on time-frequency analysis, image processing and neural networks. The method comprises the steps of firstly carrying out time-frequency analysis on LPI radar signals to obtain time-frequency distribution images, then preprocessing the time-frequency images by using an image processing method, and finally carrying out identification classification on the processed images by using an RBF neural network.
In summary, the existing signal identification and separation technology widely adopts the neural network, and the introduction of the neural network puts high requirements on a hardware module for signal processing, so that huge calculation force is required to be consumed, and the speed is low. Secondly, the recognition accuracy rate achieved by the technology under the electromagnetic environment of the aliasing of various modulation signals is very limited, and the technology is difficult to adapt to complex and changeable electromagnetic environments.
Disclosure of Invention
In view of the above-mentioned shortcomings of the prior art, the method for identifying the modulation type of the aliasing signal based on time-frequency analysis and constellation diagram analysis designed by the present invention can quickly and accurately identify signals of different modulation modes from the aliasing signal under the condition of low computational effort investment, and provides a quantitative measurement index for channel quality based on parameter fitting and small-range variance estimation.
In order to achieve the above object, the technical scheme of the present invention is as follows, and an aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis, the method comprises the following steps:
step 1: identifying the analog modulation type in the aliasing signal by using a method of combining time-frequency analysis with morphological fitting;
Step 2: the method of constellation analysis combined with edge detection is used to identify the type of digital modulation in the aliased signal. In the step 1, for the analog modulation type in the aliasing signal, a morphological fitting method is used to process a time-frequency distribution diagram obtained by time-frequency analysis, so as to identify the analog modulation type in the aliasing signal.
In the step 2, for the digital modulation type in the aliasing signal, an edge detection method is used to process the constellation diagram obtained by the constellation diagram analysis, so as to identify the digital modulation type in the aliasing signal.
The morphological fitting algorithm in the step 1 specifically comprises the following steps of processing a time-frequency distribution diagram by using the morphological fitting algorithm to sort out an analog modulation signal, firstly, carrying out smoothing processing on the time-frequency distribution gray level diagram to extract a gray level vector, then, searching a signal peak value by using a sliding window, and finally, carrying out fitting by using a least square method to finish the identification of the analog modulation signal.
In step 1, the time-frequency diagram obtained by time-frequency analysis of the signal is essentially a gray-scale diagram, in order to smooth the gray-scale diagram and eliminate the influence of noise in the image, the method is adopted in which the time-frequency diagram is cut into a plurality of horizontal stripes by taking every 5 pixel points as a unit in the longitudinal direction, then the horizontal stripe image is calculated, the gray values of the 5 pixels in the longitudinal direction are added to obtain a one-dimensional gray-scale vector, the influence of noise in the image can be reduced as much as possible by the processing in the step, and the calculation amount is reduced,
The original gray matrix is as follows:
the gray vectors extracted by calculation are as follows:
Wherein:
G11＝g11+g21+g31+g41+g51 (3)
G12＝g12+g22+g32+g42+g52 (4)
G13＝g13+g23+g33+g43+g51 (5)
After the processed gray matrix is obtained, a sliding window is used for searching a signal peak value, the sliding window is set to be a triangular window with the width of 7 pixels, the sliding window is convolved with each gray vector at the rate of moving one unit pixel each time, and finally a group of new feature vectors are obtained; after this operation is performed on all the gray vectors, the peaks of the respective feature vectors are divided into respective groups, each group corresponding to a set of curve midpoints of a modulation scheme in the time-frequency diagram. The following formula represents a triangular window of 7 pixels in width:
sequentially rolling the sliding window and the gray level vector extracted from the gray level map to obtain a new feature vector, taking the first row feature vector as an example, rolling the sliding window to obtain the following feature vector
[G′11 G′12 G′13…G′1N] (8)
Wherein the method comprises the steps of
N＝n-6# (11)
The set of points corresponding to each curve has been obtained in the above process, and then the type of curve is determined by analyzing and processing the points in each group.
In step 1, in order to achieve the purpose of judging the curve type, a least square method is first used to fit a straight line, the least square method is used to calculate a group of data conforming to the relationship of y=a+bx, the best a and b are calculated,
First, let the linear expression be:
y＝a+bx (12)
For a set of data (x i,yi) satisfying the above relationship, assuming that the error of the argument x i is ignored, under the same x i, there is the following relationship between the measured point y i and the deviation d i of the corresponding point a+bx i on the straight line:
then a and b are optimal when d 1＝d2＝…dn =0, but the actual measured data cannot be taken to be d 1＝d2＝…dn =0, so that only d 1 2+d2 2+…dn 2 is considered to be the smallest for a and b, and a and b are the best;
Order the
D, respectively solving first-order partial derivatives of a and b:
And then, second order partial derivatives are obtained:
Obviously
And (3) meeting the minimum condition, and enabling the first-order partial derivative to be zero:
average value is introduced:
Then:
and (3) solving to obtain:
Substituting a and b into an equation y=a+bx, namely, a linear equation obtained by least square fitting;
Will be As an evaluation criterion of the fitting effect,/>The smaller the fitting effect is, the better.
In step 2, each cluster in the graph is filled through an opening operation to prepare for subsequent edge detection, then the largest edge in the graph is extracted through a Sobel edge detection algorithm, and the type of the constellation diagram can be judged according to the number and the positions of the extracted edges, so that the signal is identified.
In the step 2, the etching operation is firstly carried out, then the expansion operation is carried out, the bright area of the small area in the image is replaced by the dark area, and the blank spots in each cluster are replaced by black, so that each cluster is more obvious in appearance, and preparation is provided for subsequent edge detection.
In step 2, the Sobel operator is used for edge detection,
The Sobel operator can be used for calculating the gray approximation of the brightness function of an image, and generates a gray vector or normal vector corresponding to an operation point, wherein the gray vector or normal vector comprises two groups of 3×3 matrixes G x 'and G y', G x 'is a matrix formed by carrying out plane convolution on the operator and the image transversely, G y' is a matrix formed by carrying out plane convolution on the operator and the image longitudinally, the brightness difference approximation of the transverse and longitudinal directions in the image can be obtained through convolution on the two directions, and the formula is as follows assuming that A is an original image, and G x、Gy is the gray value of the image after operation:
Gx＝Gx‘*A (42)
Gy＝Gy‘*A (43)
the gray scale of each point in the graph can be expressed as follows:
When a point G in the graph is greater than a certain threshold, the point is considered as an edge point in the graph, all edges in the graph can be obtained by applying the operator to the whole graph, after detecting the edges of the constellation graph, only the largest edge or the outermost edge in the graph is required to be extracted, the number of the edges is calculated, the number of clusters in the constellation graph can be determined according to the coordinate positions of the edges, and the arrangement mode of the clusters can be determined, so that the modulation mode of signals can be determined. The aliasing signals comprise two analog modulation signals of LFM and SFM and seven digital modulation signals of BPSK, QPSK, 8PSK, 16QAM, 64QAM, 256QAM and 1024 QAM.
The linear fitting equation and the fitting standard deviation of each curve in the time-frequency diagram can be obtained according to the previous work, and the modulation modes can be classified according to the results.
Previously, the least squares method did a straight line fit, so the fit to the cosine curve was the worst,Also maximum, the value actually calculated according to the previous algorithm can also obtain the fitting cosine curve/>The value is the largest, so by first setting the pair/>To determine whether it belongs to the cosine modulated signal. The judgment of the remaining two types of modulation modes is carried out by judging the slope of the fitting straight line. The slope of the third type of curve may not be zero due to the error in actually acquiring the data, but is never quite large, so that by setting the threshold value for b, it can be judged whether it belongs to the chirp signal or the phase shift modulation signal and the quadrature amplitude modulation signal.
Constellation diagram analysis based on Sobel edge detection is mainly divided into two parts of constellation diagram drawing and Sobel edge detection algorithm.
In the foregoing, features are extracted from the time-frequency diagram by using an image processing method to successfully classify the chirp signals and the sinusoidal frequency-modulated signals, but the digital modulation signals of phase modulation and amplitude modulation cannot be identified and classified, and the two signals are identified and classified by adopting a constellation diagram analysis method.
In general, for a modulated signal, the following equation may be used:
Wherein the method comprises the steps of
N＝1,2,…,N0 (32)
m＝1,2,…,m0 (33)
n＝1,2,…,n0 (34)
k＝1,2,…,k0 (35)
In the above formula, g (T) is a low-pass pulse waveform, and for convenience of analysis, g (T) =1, 0< T is less than or equal to T. Where N 0 denotes the number of message sequences, typically an integer power of 2. For the transmission of these N 0 message sequences, which are mapped onto the frequency, amplitude and phase of the modulated signal, respectively, then there are:
N0＝m0×n0×k0 (36)
then the different mapping modes of frequency, amplitude and phase represent different modulation modes, and two modes of classification are:
1) Phase shift modulated signal: i.e. f n and A m remain unchanged, only change I.e. m 0＝n0＝1,k0＝N0.
2) Quadrature amplitude modulation signal: i.e. only f n remains unchanged, changes A m andI.e., n 0＝1,N0＝m0×k0.
Expanding the general formula of the modulated signal:
the following set of basis vectors is selected by spatial theory:
in the above The modulated signal may be based on the above-described vector and represented by the following vector in signal space.
The constellation is analyzed using the Sobel edge detection algorithm as follows.
The key point of constellation diagram identification is that the number of clusters in the diagram and the arrangement mode thereof are defined, and the digital modulation mode belonging to a certain phase modulation and amplitude modulation can be judged. This can be achieved by detecting the edges of clusters in the constellation. Specifically, each cluster in the graph is filled through an opening operation to prepare for subsequent edge detection, then the largest edge in the graph is extracted through a Sobel edge detection algorithm, and the type of the constellation diagram can be judged according to the number and the positions of the extracted edges, so that signals are identified.
First, an opening operation, which is a morphologically important operation, is introduced, which is a combination of expansion and etching operations. The expansion is to connect adjacent elements in the image, and the erosion is to divide the individual elements in the image. The opening operation is that the etching operation is firstly carried out and then the expansion operation is carried out, so that the bright area of the small area in the image can be replaced by the dark area, and the blank spots in each cluster can be obviously seen to be replaced by black, so that each cluster is more obvious and preparation is provided for subsequent edge detection.
The Sobel operator is an operator very commonly used in the edge detection technology, and can calculate the gray approximation value of the image brightness function, and can generate a gray vector or normal vector corresponding to an operation point.
After the edges of the constellation diagram are detected, the maximum edges or the outermost edges in the diagram are only needed to be extracted, the number of the edges or the outermost edges is calculated, the number of clusters in the constellation diagram and the arrangement mode can be determined according to the coordinate positions of the edges or the outermost edges, and the modulation mode of the signals can be determined.
Compared with the prior art, the invention has the following advantages:
(1) Compared with the common algorithm for carrying out signal type identification by utilizing a neural network, the method bypasses the network training process by using a large amount of data, directly carries out modulation signal type identification by using an image processing technology, and further has the advantages of simplicity, light weight and small calculation force requirement;
(2) The method is directed to signals with a plurality of modulation types including LFM, SFM, QPSK, BPSK, 8PSK, 16QAM, 64QAM, 256QAM and 1024QAM which are mutually overlapped, forms quick cognition on surrounding complex spectrum environment on the basis of no neural network,
(3) The method can rapidly sort various signals including radar and communication signals, and has strong practical value in special scenes such as wideband reconfigurable anti-interference, intelligent electromagnetic situation awareness and the like.
Drawings
Fig. 1 is a flow chart of a method for identifying an aliased signal modulation type based on time-frequency analysis and constellation analysis.
Fig. 2 is a time-frequency analysis flow.
Fig. 3 is a morphology fitting algorithm flow.
Fig. 4 is a constellation analysis and edge detection algorithm flow.
Fig. 5 is a comparison of the 16QAM constellation processing before and after 25dB signal-to-noise ratio.
Fig. 6 is a time-frequency distribution diagram of an aliased signal.
Fig. 7 is a time-frequency gray scale map after smooth denoising.
Fig. 8 is a constellation of various signal modulation schemes at a signal-to-noise ratio of 25 dB.
Fig. 9 is a constellation diagram after edge detection.
Detailed Description
The present application will be further described with reference to the following specific examples, which are intended to be illustrative of the application and not to be limiting of the scope of the application, since various modifications of the application, which are equivalent to those skilled in the art to which the application pertains, will fall within the scope of the application as defined in the claims appended hereto.
Example 1: referring to fig. 1-9, a method for identifying an aliasing signal modulation type based on time-frequency analysis and constellation analysis, the method comprising the steps of:
step 1: identifying the analog modulation type in the aliasing signal by using a method of combining time-frequency analysis with morphological fitting;
step 2: the method of constellation analysis combined with edge detection is used to identify the type of digital modulation in the aliased signal.
As shown in fig. 1, a time-frequency analysis method (including short-time fourier transform, morlet wavelet transform, etc.) is used to draw a time-frequency distribution diagram, and a morphological fitting algorithm is used to identify an analog modulation signal in the time-frequency distribution diagram. For the digital modulation type in the aliasing signal, a constellation diagram of the signal is drawn by using a constellation diagram analysis method, and the digital modulation signal in the constellation diagram is identified by using an edge detection algorithm.
The time-frequency analysis flow is shown in fig. 2, firstly, discretizing an aliasing signal as input, placing a sliding window with the length of N on a time domain according to the number of sampling points N, sliding the sliding window along a time axis, drawing frequency spectrums for the signal in the window, and finally splicing all the frequency spectrums to form a time-frequency distribution image.
The flow of the morphological fitting algorithm is shown in fig. 3, the morphological fitting algorithm is used for processing a time-frequency distribution diagram to sort out an analog modulation signal, firstly, smoothing is needed to be carried out on the time-frequency distribution gray scale diagram, a gray scale vector is extracted, then a sliding window is used for searching a signal peak value, and finally, the least square method is used for fitting, so that the identification of the analog modulation signal is completed.
Because the time-frequency diagram obtained by time-frequency analysis of the signals is essentially a gray-scale diagram, in order to smooth the gray-scale diagram and eliminate the influence caused by noise in the image, the adopted method is to cut the time-frequency diagram into a plurality of transverse strips by taking every 5 pixel points as a unit in the longitudinal direction, then calculate the transverse strip image and add the gray-scale values of the 5 pixels in the longitudinal direction to obtain a one-dimensional gray-scale vector. By this processing, the influence of noise in the image can be reduced as much as possible, and the calculation amount can be reduced.
The original gray matrix is as follows:
the gray vectors extracted by calculation are as follows:
Wherein:
G11＝g11+g21+g31+g41+g51 (3)
G12＝g12+g22+g32+g42+g52 (4)
G13＝g13+g23+g33+g43+g51 (5)
After the processed gray matrix is obtained, a sliding window is used for searching a signal peak value. The sliding window is set to be a triangular window of 7 pixels wide, and the sliding window is convolved with each gray scale vector at a rate of one unit pixel per movement, to finally obtain a set of new feature vectors. The purpose of this operation is to take into account the correlation between adjacent vector elements and then find the peak value of the set of feature vectors, preserving the peak value with gray values greater than the threshold, i.e. the point in the plot where the curve appears. After this operation is performed on all the gray vectors, the peaks of the respective feature vectors are divided into respective groups, each group corresponding to a set of curve midpoints of a modulation scheme in the time-frequency diagram. The following formula represents a triangular window of 7 pixels in width:
sequentially rolling the sliding window and the gray level vector extracted from the gray level map to obtain a new feature vector, taking the first row feature vector as an example, rolling the sliding window to obtain the following feature vector
[G′11 G′12 G′13…G′1N] (8)
Wherein the method comprises the steps of
N＝n-6#(11)
The set of points corresponding to each curve has been obtained in the above process, and then the type of curve is determined by analyzing and processing the points in each group. To determine the type of curve, a line is fitted by a least square method, which is a method for calculating a set of data satisfying the relationship of y=a+bx to obtain the optimal values a and b.
First, let the linear expression be:
y＝a+bx (12)
For a set of data (x i,yi) satisfying the above relationship, assuming that the error of the argument x i is negligible, at the same x i, there is the following relationship between the measured point y i and the deviation d i of the corresponding point a+bx i on the straight line:
Then a and b are optimal when d 1＝d2＝…dn =0, but the actual measured data is unlikely to make d 1＝d2＝…dn =0. Thus, only when d 1 2+d2 2+…dn 2 is considered to be the smallest for a and b, the resulting a and b are the best.
Order the
D, respectively solving first-order partial derivatives of a and b:
And then, second order partial derivatives are obtained:
Obviously
And (3) meeting the minimum condition, and enabling the first-order partial derivative to be zero:
average value is introduced:
Then:
and (3) solving to obtain:
substituting a and b into the equation y=a+bx, namely, a straight line equation obtained by least square fitting.
Will beAs an evaluation criterion of the fitting effect,/>The smaller the fitting effect is, the better.
The linear fitting equation and the fitting standard deviation of each curve in the time-frequency diagram can be obtained according to the previous work, and the modulation modes can be classified according to the results.
Previously, the least squares method did a straight line fit, so the fit to the cosine curve was the worst,Also maximum, the value actually calculated according to the previous algorithm can also obtain the fitting cosine curve/>The value is the largest, so by first setting the pair/>To determine whether it belongs to the cosine modulated signal. The judgment of the remaining two types of modulation modes is carried out by judging the slope of the fitting straight line. The slope of the third type of curve may not be zero due to the error in actually acquiring the data, but is never quite large, so that by setting the threshold value for b, it can be judged whether it belongs to the chirp signal or the phase shift modulation signal and the quadrature amplitude modulation signal.
The constellation diagram analysis based on Sobel edge detection is mainly divided into two parts of constellation diagram drawing and Sobel edge detection algorithm as shown in fig. 4.
In the foregoing, features are extracted from the time-frequency diagram by using an image processing method to successfully classify the chirp signals and the sinusoidal frequency-modulated signals, but the digital modulation signals of phase modulation and amplitude modulation cannot be identified and classified, and the two signals are identified and classified by adopting a constellation diagram analysis method.
In general, for a modulated signal, the following equation may be used:
Wherein the method comprises the steps of
N＝1,2,…,N0 (32)
m＝1,2,…,m0 (33)
n＝1,2,…,n0 (34)
k＝1,2,…,k0 (35)
In the above formula, g (T) is a low-pass pulse waveform, and for convenience of analysis, g (T) =1, 0< T is less than or equal to T. Where N 0 denotes the number of message sequences, typically an integer power of 2. For the transmission of these N 0 message sequences, which are mapped onto the frequency, amplitude and phase of the modulated signal, respectively, then there are:
N0＝m0×n0×k0 (36)
then the different mapping modes of frequency, amplitude and phase represent different modulation modes, and two modes of classification are:
1) Phase shift modulated signal: i.e. f n and A m remain unchanged, only change I.e. m 0＝n0＝1,k0＝N0.
2) Quadrature amplitude modulation signal: i.e. only f n remains unchanged, changes A m andI.e., n 0＝1,N0＝m0×k0.
Expanding the general formula of the modulated signal:
the following set of basis vectors is selected by spatial theory:
in the above The modulated signal may be based on the above-described vector and represented by the following vector in signal space.
The constellation is analyzed using the Sobel edge detection algorithm as follows.
The key point of constellation diagram identification is that the number of clusters in the diagram and the arrangement mode thereof are defined, and the digital modulation mode belonging to a certain phase modulation and amplitude modulation can be judged. This can be achieved by detecting the edges of clusters in the constellation. Specifically, each cluster in the graph is filled through an opening operation to prepare for subsequent edge detection, then the largest edge in the graph is extracted through a Sobel edge detection algorithm, and the type of the constellation diagram can be judged according to the number and the positions of the extracted edges, so that signals are identified.
First, an opening operation, which is a morphologically important operation, is introduced, which is a combination of expansion and etching operations. The expansion is to connect adjacent elements in the image, and the erosion is to divide the individual elements in the image. The opening operation is that the etching operation is performed first and then the expansion operation is performed, so that the bright area of the small area in the image can be replaced by the dark area, as shown in fig. 5, wherein a) in fig. 5 is a constellation diagram of the 16QAM modulation signal, b) in fig. 5 is a constellation diagram after the opening operation is performed, and the effect of opening the operation to the opening operation can be obvious is that the blank spots in each cluster are replaced by black, so that each cluster is more obvious in appearance and preparation is provided for subsequent edge detection.
The Sobel operator is an operator very commonly used in the edge detection technology, and can calculate the gray approximation value of the image brightness function, and can generate a gray vector or normal vector corresponding to an operation point.
The Sobel operator comprises two groups of 3×3 matrixes G x 'and G y', wherein G x 'is a matrix of plane convolution of the operator and the image in the transverse direction, G y' is a matrix of plane convolution of the operator and the image in the longitudinal direction, and the approximate values of the brightness difference in the transverse direction and the longitudinal direction in the image can be obtained through convolution in the two directions, and the formula is as follows assuming that a is an original image, and G x、Gy is the gray value of the image after operation:
Gx＝Gx‘*A (42)
Gy＝Gy‘*A (43)
the gray scale of each point in the graph can be expressed by the following formula
When the G of a point in the graph is greater than a certain threshold, the point is considered as an edge point in the graph, and all edges in the image can be obtained by applying the operator to the whole graph.
After the edges of the constellation diagram are detected, the maximum edges or the outermost edges in the diagram are only needed to be extracted, the number of the edges or the outermost edges is calculated, the number of clusters in the constellation diagram and the arrangement mode can be determined according to the coordinate positions of the edges or the outermost edges, and the modulation mode of the signals can be determined.
In view of the fact that most signal recognition algorithms do not give evaluation indexes for the quality of signals after completing recognition tasks, the front end cannot be fed back, so that the front end adjusts own parameters, the purposes of reconfigurability and self-adaption are achieved, the accuracy of signal receiving is improved, and noise is reduced. The signal quality detection method based on parameter fitting and small-range variance estimation is provided, and a quantization parameter related to signal quality can be provided, so that the radio frequency receiving front end is guided to carry out self-adaptive adjustment, and the signal quality is improved.
Specific examples:
first exemplary embodiment of the present invention:
an embodiment of an aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis is provided in the invention:
the input signal is composed of one of digital modulation schemes such as linear frequency modulation LFM, sinusoidal frequency modulation SFM, BPSK, QPSK, 8PSK, and 16 QAM. The time-frequency distribution diagram is plotted using short-time fourier transform (DFT) on the input signal at a signal-to-noise ratio of 25dB as shown in fig. 6:
It is easy to see that the signal is divided into three parts in the time-frequency domain, the three parts can be distinguished by a morphology fitting method, the two analog modulation types are separated, and the analog modulation type and the digital modulation type are separated.
Firstly, carrying out smooth denoising treatment on signals, taking 5 pixels as a unit in the longitudinal direction, cutting a time-frequency diagram into a plurality of horizontal bars, and superposing the five longitudinal pixel values into a one-dimensional gray value vector. The processed image is shown in fig. 7.
A triangular sliding window of seven pixels is used, which is convolved with each gray vector at a rate of one pixel per movement to obtain a new set of feature vectors. Searching the peak value of the group of feature vectors, and reserving the peak value larger than the threshold value to achieve the purpose of grabbing the bright spots on the curve in the time-frequency distribution diagram. After the same processing is carried out on all the gray vectors, the peak values of the characteristic vectors are divided into groups, and each group corresponds to a set of curve midpoints of a modulation mode in a time-frequency diagram.
A line fitting was performed using the least square method. The results are shown in FIG. 7.
The following expressions correspond to the order of signals from left to right in fig. 7:
y1＝0.270193x1+13.7078 (45)
y2＝65 (47)
y3＝-0.132145x3+115.846 (49)
According to the set threshold value, the left signal is judged to be linear frequency modulation LFM, the right signal is judged to be sinusoidal frequency modulation, and the central signal is judged to be one of several digital modulation modes. The following gives a method for identifying the relevant signals in combination with a method for constellation analysis.
For BPSK, QPSK, 8PSK and 16QAM modulation signals in digital modulation, a constellation diagram analysis and Sobel edge detection method is adopted for identification. For the above signal, the constellation diagram is shown in fig. 8 with a signal-to-noise ratio of 25 dB.
In order to automatically identify the signals from the constellation, the constellation is analyzed using a Sobel edge detection algorithm. Firstly, each cluster in the graph is filled through an opening operation, then the maximum edge in the graph is extracted through a Sobel edge detection algorithm, and the type of the constellation diagram can be judged according to the number and the positions of the extracted edges. The constellation after edge detection is drawn as shown in fig. 9.
The number of clusters in the graph can be calculated to be 16, the arrangement mode accords with the 16QAM characteristic, and the signal is determined to be a 16QAM modulation mode. The same applies to other modulation schemes.
It should be noted that the above-mentioned embodiments are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention, and the equivalent substitutions or alternatives made on the basis of the above-mentioned technical solutions are all included in the scope of the present invention.
Claims (8)
1. An aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis, which is characterized by comprising the following steps:
step 1: identifying the analog modulation type in the aliasing signal by using a method of combining time-frequency analysis with morphological fitting;
Step 2: identifying a digital modulation type in the aliasing signal by using a constellation diagram analysis and edge detection combined method; in step 1, the time-frequency diagram obtained by performing time-frequency analysis on the signal is essentially a gray-scale diagram, in order to smooth the gray-scale diagram and eliminate the influence of noise in the image, the method is adopted in which the time-frequency diagram is cut into a plurality of horizontal stripes by taking every 5 pixel points as a unit in the longitudinal direction, then the horizontal stripe image is calculated, and the gray values of the 5 pixels in the longitudinal direction are added to obtain a one-dimensional gray-scale vector, and by the processing in this step, the influence of noise in the image can be reduced and the calculation amount is reduced,
The original gray matrix is as follows:
the gray vectors extracted by calculation are as follows:
Wherein:
G11＝g11+g21+g31+g41+g51 (3)
G12＝g12+g22+g32+g42+g52 (4)
G13＝g13+g23+g33+g43+g51 (5)
After the processed gray matrix is obtained, a sliding window is used for searching a signal peak value, the sliding window is set to be a triangular window with the width of 7 pixels, the sliding window is convolved with each gray vector at the rate of moving one unit pixel each time, and finally a group of new feature vectors are obtained; after this operation is performed on all the gray vectors, the peaks of the feature vectors are divided into groups, each group corresponds to a set of curve midpoints of a modulation mode in the time-frequency diagram, and the following formula represents a triangular window with 7 pixels in width:
sequentially rolling the sliding window and the gray level vector extracted from the gray level map to obtain a new feature vector, taking the first row feature vector as an example, rolling the sliding window to obtain the following feature vector
[ G ′ 11G′ 12G′ 13…G′ 1N ] (8) wherein
N＝n-6 (11)
The set of points corresponding to each curve is obtained in the process, and then the type of the curve is judged by analyzing and processing the points in each group;
In step 1, in order to achieve the purpose of judging the curve type, a least square method is first used to fit a straight line, the least square method is used to calculate a group of data conforming to the relationship of y=a+bx, the best a and b are calculated,
First, let the linear expression be:
y＝a+bx (12)
For a set of data (x i,yi) satisfying the above relationship, assuming that the error of the argument x i is ignored, under the same x i, there is the following relationship between the measured point y i and the deviation d i of the corresponding point a+bx i on the straight line:
then a and b are optimal when d 1＝d2＝…dn =0, but the actual measured data cannot be taken to be d 1＝d2＝…dn =0, so that only d 1 2+d2 2+…dn 2 is considered to be the smallest for a and b, and a and b are the best;
Order the
D, respectively solving first-order partial derivatives of a and b:
And then, second order partial derivatives are obtained:
Obviously
And (3) meeting the minimum condition, and enabling the first-order partial derivative to be zero:
average value is introduced:
Then:
and (3) solving to obtain:
Substituting a and b into an equation y=a+bx, namely, a linear equation obtained by least square fitting;
Will be As an evaluation criterion of the fitting effect,/>The smaller the fitting effect is, the better.
2. The method for identifying the modulation type of the aliasing signal based on the time-frequency analysis and the constellation diagram analysis according to claim 1, wherein in the step 1, for the analog modulation type in the aliasing signal, a morphological fitting method is used to process a time-frequency distribution diagram obtained by the time-frequency analysis, so as to identify the analog modulation type in the aliasing signal.
3. The method for identifying the modulation type of the aliasing signal based on the time-frequency analysis and the constellation analysis according to claim 1, wherein in the step 2, the constellation obtained by the constellation analysis is processed by using an edge detection method for the digital modulation type in the aliasing signal, so as to identify the digital modulation type in the aliasing signal.
4. The method for identifying the modulation type of the aliasing signal based on the time-frequency analysis and the constellation diagram analysis according to claim 2, wherein the morphological fitting algorithm in the step 1 is specifically as follows, the analog modulation signal is sorted out by using the time-frequency distribution diagram processed by the morphological fitting algorithm, firstly, a gray level vector is extracted by smoothing the time-frequency distribution gray level diagram, then, a signal peak value is found by using a sliding window, and finally, the analog modulation signal is identified by using a least square method.
5. The method for identifying the modulation type of the aliasing signal based on time-frequency analysis and constellation diagram analysis according to claim 1, wherein in step 2, each cluster in the diagram is filled by an open operation to prepare for subsequent edge detection, then the largest edge in the diagram is extracted by a Sobel edge detection algorithm, and the type of the constellation diagram can be judged according to the number and the position of the extracted edges, so as to identify the signal.
6. The method for identifying an aliasing signal modulation type based on time-frequency analysis and constellation analysis according to claim 5, wherein in step 2, the etching operation is performed first, then the expanding operation is performed, the bright areas of the small areas in the image are replaced with dark areas, and the blank spots in each cluster are replaced with black, so that each cluster appears more obvious, and preparation is made for subsequent edge detection.
7. The method for identifying aliasing signal modulation type based on time-frequency analysis and constellation analysis of claim 6 wherein in step 2, edge detection is performed using a Sobel operator which is used to calculate a gray approximation of the image luminance function which produces a gray vector or normal "" corresponding to the operation point'
The vector comprises two groups of 3×3 matrixes G x and G y, wherein G x is a matrix in which the operator and the image are subjected to plane convolution in the transverse direction, G y' is a matrix in which the operator and the image are subjected to plane convolution in the longitudinal direction, and the approximate values of the brightness difference in the transverse direction and the longitudinal direction in the image can be obtained through convolution in the two directions, and the formula is as follows assuming that a is an original image and G x、Gy is an image gray value after operation:
Gx＝Gx‘*A (42)
Gy＝Gy‘*A (43)
the gray scale of each point in the graph can be expressed as follows:
when a point G in the image is larger than a certain threshold, the point G is considered as an edge point in the image, all edges in the image can be obtained by applying the operator to the whole image, after detecting the edges of the constellation image, only the largest edge or the outermost edge in the image is required to be extracted, the number of the edges is calculated, the number of clusters in the constellation image can be determined according to the coordinate positions of the edges, and the arrangement mode can be determined, so that the modulation mode of signals can be determined.
8. The method for identifying the modulation type of the aliasing signal based on the time-frequency analysis and the constellation analysis according to claim 1, wherein the aliasing signal comprises two analog modulation signals of LFM and SFM and seven digital modulation signals of BPSK, QPSK, 8PSK, 16QAM, 64QAM, 256QAM and 1024 QAM.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110597435.9A CN113536897B (en) | 2021-05-31 | 2021-05-31 | Aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110597435.9A CN113536897B (en) | 2021-05-31 | 2021-05-31 | Aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113536897A CN113536897A (en) | 2021-10-22 |
| CN113536897B true CN113536897B (en) | 2024-04-30 |
Family
ID=78095050
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110597435.9A Active CN113536897B (en) | 2021-05-31 | 2021-05-31 | Aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113536897B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102508025A (en) * | 2010-05-13 | 2012-06-20 | 特克特朗尼克公司 | Signal recognition and triggering using computer vision techniques |
| KR101910540B1 (en) * | 2018-03-14 | 2018-10-22 | 국방과학연구소 | Apparatus and method for recognizing radar waveform using time-frequency analysis and neural network |
| CN108875706A (en) * | 2018-07-18 | 2018-11-23 | 中国海洋大学 | The ocean structure Time-Frequency Analysis Method collected based on sliding average and energy |
| CN109885805A (en) * | 2019-01-29 | 2019-06-14 | 南京工业职业技术学院 | A kind of instantaneous Frequency Estimation method of multi -components non-stationary signal |
-
2021
- 2021-05-31 CN CN202110597435.9A patent/CN113536897B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102508025A (en) * | 2010-05-13 | 2012-06-20 | 特克特朗尼克公司 | Signal recognition and triggering using computer vision techniques |
| KR101910540B1 (en) * | 2018-03-14 | 2018-10-22 | 국방과학연구소 | Apparatus and method for recognizing radar waveform using time-frequency analysis and neural network |
| CN108875706A (en) * | 2018-07-18 | 2018-11-23 | 中国海洋大学 | The ocean structure Time-Frequency Analysis Method collected based on sliding average and energy |
| CN109885805A (en) * | 2019-01-29 | 2019-06-14 | 南京工业职业技术学院 | A kind of instantaneous Frequency Estimation method of multi -components non-stationary signal |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113536897A (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lopes et al. | Adaptive speckle filters and scene heterogeneity | |
| US11468273B2 (en) | Systems and methods for detecting and classifying anomalous features in one-dimensional data | |
| CN110400267A (en) | A kind of preprocess method based on inspection image | |
| CN102819827B (en) | Self-adaption moment matching stripe noise removing method based on gray-level segmentation | |
| Kansal et al. | Improved color attenuation prior based image de-fogging technique | |
| Fontes et al. | Performance evaluation of the correntropy coefficient in automatic modulation classification | |
| CN111885624B (en) | Signal noise floor estimation method based on deep learning | |
| CN112070717B (en) | Power transmission line icing thickness detection method based on image processing | |
| CN110187313B (en) | Radar signal sorting and identifying method and device based on fractional order Fourier transform | |
| Sun et al. | An adaptive segmentation method combining MSRCR and mean shift algorithm with K-means correction of green apples in natural environment | |
| Zhu et al. | Automatic recognition of radar signals based on time-frequency image character | |
| CN114025379B (en) | Broadband multi-signal detection method, device and equipment | |
| KR100393071B1 (en) | Apparatus and method for histogram equlization of image | |
| CN113312996B (en) | Detection and identification method for aliasing short-wave communication signals | |
| CN113298846A (en) | Intelligent interference detection method based on time-frequency semantic perception | |
| CN111585662B (en) | Classification identification and parameter estimation method and system for phase modulation signal | |
| CN116359851A (en) | Radar active interference detection and identification method and device based on converged network | |
| CN109472788A (en) | A kind of scar detection method on airplane riveting surface | |
| CN114170103A (en) | Electrical equipment infrared image enhancement method | |
| CN111612130A (en) | Frequency shift keying communication signal modulation mode identification method | |
| CN113536897B (en) | Aliasing signal modulation type identification method based on time-frequency analysis and constellation diagram analysis | |
| Ju et al. | A novel fully convolutional network based on marker-controlled watershed segmentation algorithm for industrial soot robot target segmentation | |
| Zhengzhou et al. | Gray-scale edge detection and image segmentation algorithm based on mean shift | |
| CN114578384B (en) | Self-adaptive constant false alarm detection method for laser atmospheric system | |
| Xiao et al. | Underwater image classification based on image enhancement and information quality evaluation |
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 |