CN116582195B - Unmanned aerial vehicle signal spectrum identification method based on artificial intelligence - Google Patents
Unmanned aerial vehicle signal spectrum identification method based on artificial intelligence Download PDFInfo
- Publication number
- CN116582195B CN116582195B CN202310691995.XA CN202310691995A CN116582195B CN 116582195 B CN116582195 B CN 116582195B CN 202310691995 A CN202310691995 A CN 202310691995A CN 116582195 B CN116582195 B CN 116582195B
- Authority
- CN
- China
- Prior art keywords
- signal
- frequency
- ltp
- local
- mode
- 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
- 238000001228 spectrum Methods 0.000 title claims abstract description 57
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 13
- 238000002360 preparation method Methods 0.000 claims abstract description 31
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 238000000605 extraction Methods 0.000 claims description 76
- 238000005070 sampling Methods 0.000 claims description 41
- 238000001914 filtration Methods 0.000 claims description 25
- 230000003068 static effect Effects 0.000 claims description 19
- 238000012937 correction Methods 0.000 claims description 15
- 230000004927 fusion Effects 0.000 claims description 11
- 239000000523 sample Substances 0.000 claims description 10
- 230000005540 biological transmission Effects 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 7
- 238000012300 Sequence Analysis Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000010183 spectrum analysis Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 210000005069 ears Anatomy 0.000 claims description 3
- 238000009432 framing Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 239000013074 reference sample Substances 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 2
- 125000003275 alpha amino acid group Chemical group 0.000 claims 1
- 150000001875 compounds Chemical class 0.000 claims 1
- 230000010365 information processing Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000002457 bidirectional effect Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/0082—Monitoring; Testing using service channels; using auxiliary channels
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Electromagnetism (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an unmanned aerial vehicle signal spectrum recognition method based on artificial intelligence, which relates to the technical field of data information processing and solves the problems that the unmanned aerial vehicle signal spectrum recognition capability is lagged, and an unmanned aerial vehicle signal spectrum recognition algorithm comprises the following steps: receiving a signal; storing the signal; preprocessing signals; digitizing the signal; demodulating the signals; and (5) signal classification. The improved local three-value mode LTP algorithm is adopted to obtain a preparation characteristic value, noise influence is reduced, edge adaptability is enhanced, the method of converting the LTP characteristic into the LFBF characteristic is adopted to increase robustness and expressive of the spectrum characteristic, the time domain differential entropy TDE algorithm and the LFBF algorithm are adopted to be fused to obtain a more comprehensive unmanned plane signal characteristic, the improved K-means algorithm is adopted to intelligently demodulate signals, the signals are intelligently classified and identified through the classification module according to the signal frequency band, the power, the modulation mode and the receiving and transmitting mode, and the unmanned plane signal spectrum identification capability is greatly improved.
Description
Technical Field
The invention relates to data information processing, in particular to an unmanned aerial vehicle signal spectrum identification method based on artificial intelligence.
Background
As unmanned aerial vehicles are increasingly used, potential safety hazards caused by unmanned aerial vehicles are also increasingly concerned. Wherein the drone may collide with an organic person or be used as a terrorist attack and spy. Therefore, the need for monitoring and identifying drone signals has increased. One of the key technologies of unmanned aerial vehicle identification and tracking is to analyze and identify the signal spectrum thereof. The unmanned aerial vehicle signal spectrum recognition algorithm refers to a method for recognizing the signal spectrum received by the unmanned aerial vehicle. In unmanned aerial vehicle applications, signal spectrum recognition algorithms are very important, which can help unmanned aerial vehicle drivers determine the position and shape of objects and acquire useful video and image information. Along with the continuous development of unmanned aerial vehicle technology and the continuous expansion of application scene, traditional unmanned aerial vehicle signal frequency spectrum recognition algorithm can't satisfy market's demand to unmanned aerial vehicle signal's discernment rate, and the influence that factors such as different geographical environment, equipment model, interference factor led to the fact the identifiable degree of unmanned aerial vehicle signal is considered comprehensively, and the market is paid more attention to the compatibility to signal spectrum diversity. In addition, the traditional unmanned aerial vehicle signal spectrum recognition algorithm cannot intelligently demodulate various unmanned aerial vehicle signals. In the prior art, the unmanned aerial vehicle carries out preprocessing on received signals, converts the signals received by the unmanned aerial vehicle into electric signals, and carries out filtering and denoising processing so as to improve the quality and reliability of the signals. The signal spectrum is then analyzed, and the signal is decomposed into different frequency components using spectral analysis techniques. For example, in image analysis, convolutional Neural Networks (CNNs) may be used to identify components of different colors and luminance frequencies. However, the signal spectrum recognition capability of the unmanned aerial vehicle in the prior art is relatively poor, when the characteristic information of each component is extracted, for example, the characteristic information such as color, brightness, kurtosis, valley degree and the like is relatively lagged, and when the extracted characteristic information is used for training a machine learning model, the recognition capability of different data information models is insufficient. How to improve unmanned aerial vehicle signal frequency spectrum recognition ability, the control unmanned aerial vehicle signal frequency spectrum recognition rule still is the technical problem that needs to be solved urgently.
Disclosure of Invention
Aiming at the defects of the technology, the invention discloses an unmanned aerial vehicle signal spectrum recognition method based on artificial intelligence, which adopts an improved local three-value mode LTP algorithm to obtain a preparation characteristic value in a pixel point neighborhood by selecting random sampling points for pixel value comparison, more comprehensively describes the characteristics of a spectrum image, reduces the influence of noise factors, adopts a mode of comparing the magnitude relation between the preparation characteristic value of a central pixel point and the preparation characteristic value of a symmetrical pixel point to obtain LTP characteristics, has better adaptability to the edge of the spectrum image, adopts a method of converting the LTP characteristics into LFBF characteristics, increases the robustness and the expressive nature of the spectrum characteristics, adopts a mode of dividing an initial signal into local frequency bands with uneven frequency distribution and even frequency bands with even frequency distribution, and respectively calculates the local frequency band complete characteristics and the frequency bands with even frequency distribution formed by the LFBF characteristics by a time domain differential entropy TDE algorithm to obtain more comprehensive unmanned aerial vehicle signal characteristics, adopts a demodulation mode required by the intelligent recognition signal of the improved K-means, and carries out automatic demodulation to obtain demodulated digital signals, and realizes the intelligent recognition of the intelligent signal by a high-level signal classification module according to the signal classification signal, the initial signal, the power signal classification mode and the intelligent signal recognition method.
In order to achieve the technical effects, the invention adopts the following technical scheme:
an unmanned aerial vehicle signal spectrum identification method based on artificial intelligence comprises the following steps:
step one, receiving unmanned aerial vehicle signals through a three-in-one antenna;
step two, storing the received original signals to a storage module;
the storage module is used for storing the received original signals sent by the unmanned aerial vehicle and used for carrying out data processing subsequently;
step three, preprocessing the signals through a data preprocessing module, and extracting signal characteristics;
the data preprocessing module carries out filtering, positive frequency and feature extraction on the signals, so that the signals are convenient to identify and analyze;
the data preprocessing module comprises a filtering submodule, a frequency correction submodule and a characteristic extraction submodule, wherein the filtering submodule adopts a band-pass filter for filtering out-of-band noise of the discretized digital signal, the frequency correction submodule is used for carrying out frequency estimation and frequency compensation on the received signal, the characteristic extraction submodule adopts a method of combining a local frequency basis function LFBF algorithm, an improved local three-value mode LTP algorithm and a time domain differential entropy TDE algorithm to extract key characteristics in the signal for subsequent spectrum analysis, the output end of the filtering submodule is connected with the input end of the frequency correction submodule, and the output end of the frequency correction submodule is connected with the input end of the characteristic extraction submodule;
Step four, confirming the signal type of the original signal through a digitizing module, and digitizing the analog signal;
the digitizing module is used for ensuring that all received original signals of the unmanned aerial vehicle are converted into digital signals;
the digitizing module comprises a type judging submodule and an A/D converter, the type judging submodule is used for judging the signal type of an original signal, the signal type comprises a digital signal and an analog signal, the analog signal is sent to the A/D converter, the A/D converter carries out time discretization and amplitude discretization on the analog signal, and the output end of the judging submodule is connected with the input end of the A/D converter;
step five, demodulating the modulated digital signal through a demodulation module;
the demodulation module is used for converting the modulated digital signals and the analog signals into demodulated digital signals;
the demodulation module comprises a modulation judging sub-module and a self-demodulation model, wherein the modulation judging sub-module is used for judging whether a digital signal is modulated or not, the modulated digital signal is sent to the self-demodulation model, the self-demodulation model adopts a demodulation mode required by an improved K-means algorithm to confirm the signal and carries out demodulation operation on the signal, and the output end of the judging sub-module is connected with the input end of the self-demodulation model;
Step six, classifying the demodulated digital signals;
the classification module classifies the demodulated digital signals according to the signal frequency band, the signal power, the initial signal modulation mode and the signal receiving and transmitting mode.
As a further technical scheme of the invention, the frequency correction submodule comprises a frequency compensation system and an estimation system, wherein the frequency compensation system is used for carrying out fast fourier transformation on a received signal to realize frequency compensation, the estimation system is used for estimating the frequency of the received signal and reducing the influence of frequency deviation on demodulation performance, and the output end of the frequency compensation system is connected with the input end of the estimation system.
As a further technical scheme of the invention, the characteristic extraction submodule comprises a frequency judging submodule, a local frequency extraction model, a time domain frequency extraction model and a fusion submodule, wherein the frequency judging submodule is used for judging whether the frequency of a frequency band in an initial signal is uniform or not, extracting the local frequency band with nonuniform frequency facilitates subsequent signal analysis, the local frequency extraction model adopts a Local Frequency Basis Function (LFBF) algorithm and an improved local three-value mode (LTP) algorithm, effectively extracts characteristic information of the local nonuniform frequency in an image transmission signal to obtain a local frequency band static image characteristic, the time domain frequency extraction model adopts a time domain differential entropy (TDE) algorithm, calculates time differential entropy and differential entropy of the signal in a time domain, extracts rule information of time domain change of the signal, and the fusion submodule is used for fusing rule information of time domain change of the signal with the uniform frequency band into an information characteristic set, the output end of the frequency judging submodule is connected with the input end of the local frequency extraction model and the time domain frequency extraction model, and the output end of the local frequency extraction model is connected with the input end of the time domain frequency extraction model, and the output end of the time domain frequency extraction model is connected with the input end of the fusion submodule.
As a further technical scheme of the invention, the characteristic extraction submodule adopts a mode of dividing an initial signal into a local frequency band with uneven frequency distribution and an even frequency band with even frequency distribution and calculating the complete characteristic of the local frequency band and the frequency band with even frequency distribution respectively through a time domain differential entropy TDE algorithm to obtain more comprehensive unmanned aerial vehicle signal characteristics, and the working mode of the characteristic extraction submodule is as follows:
c1, extracting local frequency bands with uneven frequency distribution in an initial signal, wherein the rest frequency bands are uniform frequency bands;
calculating the complete characteristics of the local frequency band, and extracting the characteristics of the local frequency band with uneven frequency distribution in the initial signal image through the local frequency extraction model to obtain the complete characteristics of the static image of the local frequency band;
performing time sequence analysis on the complete characteristics of the local frequency band, calculating time difference and difference entropy of the characteristics of the local static image on the time domain by adopting a time domain difference entropy TDE algorithm, and extracting dynamic characteristics of the local static image;
c4, carrying out time sequence analysis on the frequency bands with uniform frequency distribution, and extracting dynamic characteristics of the uniform frequency bands by adopting time difference and difference entropy of the time domain difference entropy TDE algorithm on the uniform frequency bands in the time domain;
And C5, fusing the features, and splicing the dynamic features of the static image features and the dynamic features of the uniform frequency bands according to time sequences through the fusion sub-module to obtain a complete information feature set.
As a further technical scheme of the invention, the local frequency extraction model adopts a mode of combining multi-scale features of an LTP image with the LTP features to obtain comprehensive LTP features, the accuracy of the expression of the LTP features is improved, the robustness and the expressivity of spectrum features are improved by adopting a method of converting the LTP features into LFBF features, and the local frequency extraction model works in the following modes:
step 1, framing a local frequency band, and dividing the local frequency band into frames with continuous time domains, wherein each second is 48 frames;
step 2, performing fast Fourier transform on each frame to obtain a frequency domain representation corresponding to the signal and the time domain;
step 3, mapping the frequency spectrum onto a logarithmic scale, and mapping the frequency spectrum into a coordinate system by adopting a Mel scale to obtain nonlinear mapping which can be perceived by human ears;
step 4, generating LTP images, and for each frame of spectrum image, calculating a three-value mode of difference between adjacent samples in a local area to generate corresponding LTP images;
step 5, confirming LTP features to obtain feature vectors of each frame, and extracting features of the LTP images by adopting an improved local three-value mode LTP algorithm to obtain the LTP features of each frame;
Step 6, confirming multi-scale characteristics of the LTP image, adopting a multi-scale filter to carry out filtering operation on the LTP image, and carrying out filtering on different scales on the LTP image of each frequency band to extract the multi-scale characteristics of the LTP image;
step 7, obtaining comprehensive LTP characteristics, and connecting the LTP characteristics obtained by adopting an improved local three-value mode LTP algorithm with the LTP characteristics obtained by adopting a multi-scale filter for filtering to obtain the comprehensive LTP characteristics;
step 8, obtaining LBF characteristics, carrying out normalization processing and binarization operation on the comprehensive LTP characteristics, wherein each basic mode is a characteristic, using the occurrence frequency to represent the importance degree of the characteristics, and superposing the occurrence frequency of each basic mode in an image to form an LBF characteristic vector;
and 9, obtaining the complete characteristics of the local frequency band, and splicing the LFBF characteristics of each frame into one characteristic which is the complete characteristics of the local frequency band.
As a further technical scheme of the invention, the improved local three-value mode LTP algorithm obtains a preparation characteristic value by selecting random sampling points in a pixel point neighborhood to perform pixel value comparison, describes characteristics of a signal image more comprehensively, reduces influence of noise factors, obtains LTP characteristics by comparing magnitude relation between the preparation characteristic value of a central pixel point and the preparation characteristic value of a symmetrical pixel point, has better adaptability to edges of the signal image, and comprises the following working formulas:
S1, confirming sampling points, selecting any pixel point in a spectrum image of a frame as a central pixel point, and selecting S random sampling points in the neighborhood of the central pixel point, wherein the expression of the number of the sampling points is as follows:
in the formula (1), S is the number of sampling points, r is the neighborhood radius, the number of pixel points in the neighborhood is x, and n is the total number of pixel points;
s2, coding the central pixel point, and calculating the size relation between the pixel value of the central pixel point and the pixel value of each sampling point, wherein the coding rule is as follows:
in the formula (2), X is the pixel value of the central pixel point, Y i For the pixel value of the sampling point, i is the index of the sampling point, Q α The method is based on a central pixel point coding value of a sampling point pixel, w is the average value of the sampling point pixel, and s is the standard deviation of the sampling point pixel;
s3, confirming a preparation characteristic value, connecting central pixel point codes based on sampling point pixels to obtain a binary code sequence with the length of 8, and converting binary into decimal to obtain the preparation characteristic value of the central pixel point;
s4, coding the symmetrical pixel points, confirming the symmetrical pixel points of the central pixel point according to a symmetry rule, and calculating the preparation characteristic values of the symmetrical pixel points according to the methods in S1, S2 and S3;
S5, recoding the central pixel point, and calculating the size relation between the central pixel point preparation characteristic value and the symmetrical pixel point preparation characteristic value, wherein the coding rule is as follows:
in the formula (3), A is a central pixel point preparation characteristic value, B is a symmetrical pixel point preparation characteristic value, Q β For central pixel point coding based on symmetrical pixels, Z j The pixel value j is the reference number of the sampling point in the adjacent symmetrical pixel point;
s6, confirming a characteristic value, namely connecting central pixel point codes based on symmetrical pixels to obtain a binary code sequence with the length of 8, and converting binary into decimal to obtain LTP characteristics of the central pixel point;
and S7, traversing pixel points in the spectrum image, and calculating LTP characteristics of all the pixel points in the spectrum image of one frame to obtain the LTP characteristics.
As a further technical scheme of the invention, the self-demodulation model comprises a sample library, a classification system and a modem library, wherein the sample library is used for storing modulated digital signal characteristics of a marked modulation and demodulation mode, the classification system is used for confirming a modulation mode of a modulated digital signal as a reference sample for confirming an original signal modulation mode, the modem library is used for matching a corresponding demodulation mode for the modulated digital signal according to the modulation mode confirmed by the classification system, the demodulation of the digital signal is realized, the output end of the sample library is connected with the input end of the classification system, and the output end of the classification system is connected with the input end of the modem library.
As a further technical scheme of the present invention, the improved K-means algorithm eliminates redundant data by removing orphan points, confirms a clustering center point by a fitness sorting method, increases classification accuracy of the classification system, confirms a modulation mode of a modulated digital signal to be classified by a method of calculating a distance between a digital signal feature of a modulated demodulation mode and the modulated digital signal to be classified, and the working mode of the improved K-means algorithm is as follows:
1) Confirming a K value, and fixing the K value to be 10;
2) Treating the modulated digital signal characteristic as a data point;
3) Removing the isolated points, calculating the point density of the data points, confirming that the data points with the point density lower than the density index are the isolated points, and deleting the isolated points from the data set;
4) Confirming a clustering center point, sorting data points according to the magnitude of the fitness, and selecting ten data points with the highest fitness as the clustering center point;
5) Clustering calculation, namely classifying data around a clustering center point into a data cluster to be classified according to the Euclidean distance as a radius;
6) And classifying data, namely merging the digital signal characteristics of the marked modulation and demodulation mode into the data cluster to be classified, and when the distance between the digital signal characteristics of the marked modulation and demodulation mode and the data cluster to be classified is calculated to be smaller than a distance threshold value, classifying the data cluster to be classified into the category of the digital signal characteristics of the marked modulation and demodulation mode, wherein the distance threshold value formula is as follows:
In the formula (4), P n G is the central point position of the digital signal characteristic of the marked modulation-demodulation mode m The method comprises the steps of taking a clustering center point of a data cluster to be classified, wherein k is the number of the data clusters to be classified, n is the label of the center point of the data cluster to be classified, m is the label of the clustering center point of a digital signal characteristic of a marked modulation-demodulation mode, lambda is the radius of the data cluster to be classified, v is the clustering radius of the digital signal characteristic of the marked modulation-demodulation mode, and D is a distance threshold.
The beneficial effects of the invention are as follows:
the invention discloses an unmanned aerial vehicle signal spectrum recognition method based on artificial intelligence, which is characterized in that an improved local three-value mode LTP algorithm is adopted to obtain a preliminary characteristic value in a pixel point neighborhood by selecting random sampling points for pixel value comparison, characteristics of a spectrum image are described more comprehensively, influence of noise factors is reduced, the LTP characteristic is obtained by adopting a mode of comparing magnitude relation between the preliminary characteristic value of a central pixel point and the preliminary characteristic value of a symmetrical pixel point, the edge of the spectrum image has better adaptability, the LTP characteristic is converted into an LBF characteristic, robustness and expressivity of the spectrum characteristic are increased, an initial signal is divided into a local frequency band with uneven frequency distribution and a uniform frequency band with even frequency distribution, the local frequency band complete characteristic formed by the LBF characteristic and the frequency band with even frequency distribution are calculated through a time domain differential entropy TDE algorithm, the unmanned aerial vehicle signal characteristic is obtained more comprehensively, the intelligent recognition signal is demodulated in a mode required by adopting an improved K-means, the intelligent recognition signal is demodulated automatically, the intelligent recognition signal is obtained by adopting a digital signal receiving module, and the intelligent recognition signal is classified into a high-level signal classification mode according to the intelligent signal, and the intelligent signal classification mode is realized.
Drawings
For a clearer description of embodiments of the invention or of solutions in the prior art, the drawings that are necessary for the description of the embodiments or of the prior art will be briefly described, it being apparent that the drawings in the description below are only some embodiments of the invention, from which, without inventive faculty, other drawings can be obtained for a person skilled in the art, in which:
fig. 1 is a flowchart of an unmanned aerial vehicle signal spectrum recognition method based on artificial intelligence, which is provided by the embodiment of the invention;
fig. 2 is a block diagram of an unmanned aerial vehicle signal spectrum recognition method based on artificial intelligence, which is provided by the embodiment of the invention;
FIG. 3 is a workflow diagram of a feature extraction sub-module of the present invention;
FIG. 4 is a workflow diagram of a local frequency extraction model of the present invention;
FIG. 5 is a flowchart of the operation of the modified partial three value mode LTP algorithm of the present invention.
Detailed Description
The following description of the embodiments of the present disclosure will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the disclosure. It should be understood that the description is only illustrative and is not intended to limit the scope of the invention. In addition, in the following description, descriptions of well-known structures and techniques are omitted so as not to unnecessarily obscure the present invention.
As shown in fig. 1, an unmanned aerial vehicle signal spectrum identification method based on artificial intelligence comprises the following steps:
step one, receiving unmanned aerial vehicle signals through a three-in-one antenna;
step two, storing the received original signals to a storage module;
the storage module is used for storing the received original signals sent by the unmanned aerial vehicle and used for carrying out data processing subsequently;
step three, preprocessing the signals through a data preprocessing module, and extracting signal characteristics;
the data preprocessing module carries out filtering, positive frequency and feature extraction on the signals, so that the signals are convenient to identify and analyze;
the data preprocessing module comprises a filtering submodule, a frequency correction submodule and a characteristic extraction submodule, wherein the filtering submodule adopts a band-pass filter for filtering out-of-band noise of the discretized digital signal, the frequency correction submodule is used for carrying out frequency estimation and frequency compensation on the received signal, the characteristic extraction submodule adopts a method of combining a local frequency basis function LFBF algorithm, an improved local three-value mode LTP algorithm and a time domain differential entropy TDE algorithm to extract key characteristics in the signal for subsequent spectrum analysis, the output end of the filtering submodule is connected with the input end of the frequency correction submodule, and the output end of the frequency correction submodule is connected with the input end of the characteristic extraction submodule;
Step four, confirming the signal type of the original signal through a digitizing module, and digitizing the analog signal;
the digitizing module is used for ensuring that all received original signals of the unmanned aerial vehicle are converted into digital signals;
the digitizing module comprises a type judging submodule and an A/D converter, the type judging submodule is used for judging the signal type of an original signal, the signal type comprises a digital signal and an analog signal, the analog signal is sent to the A/D converter, the A/D converter carries out time discretization and amplitude discretization on the analog signal, and the output end of the judging submodule is connected with the input end of the A/D converter;
step five, demodulating the modulated digital signal through a demodulation module;
the demodulation module is used for converting the modulated digital signals and the analog signals into demodulated digital signals;
the demodulation module comprises a modulation judging sub-module and a self-demodulation model, wherein the modulation judging sub-module is used for judging whether a digital signal is modulated or not, the modulated digital signal is sent to the self-demodulation model, the self-demodulation model adopts a demodulation mode required by an improved K-means algorithm to confirm the signal and carries out demodulation operation on the signal, and the output end of the judging sub-module is connected with the input end of the self-demodulation model;
Step six, classifying the demodulated digital signals;
the classification module classifies the demodulated digital signals according to the signal frequency band, the signal power, the initial signal modulation mode and the signal receiving and transmitting mode.
Through the above embodiment, the classifying module includes a frequency band classifying sub-module, a power classifying sub-module, a modulating classifying sub-module, a mode classifying sub-module and a summarizing sub-module, where the frequency band classifying sub-module classifies signals into regular task signals, control signals, image transmission signals and positioning signals according to different frequency bands used by the demodulated digital signals, the power classifying sub-module classifies signals into control signals, image transmission signals and telemetry signals according to different power levels of the demodulated digital signals, the modulating classifying sub-module classifies signals into control signals, telemetry signals and image transmission signals according to different modulation modes adopted by the demodulated digital signals, the mode classifying sub-module classifies signals into unidirectional signals and bidirectional signals according to the demodulated digital signals, the unidirectional signals include image transmission signals, positioning signals and telemetry signals, the bidirectional signals include regular task signals and control signals, the summarizing sub-module adopts a voting method to classify the signals into final classification results according to the frequency bands, signal power, initial signal modulation mode and signal reception mode, so that the signals are classified into control signals, telemetry signals, and classification parameters are more classified by the power classifying sub-module, the accurate mode classifying sub-module is connected with the classifying sub-module 1, and the classifying sub-module is connected with the classifying module.
TABLE 1 Classification parameter Table
As shown in table 1, the normal task signal is a non-control signal, and can be transmitted between the command receiver and the receiver, the control signal is a signal for controlling the direction, speed, altitude, etc. of the aircraft, the image transmission signal is a signal for capturing images and videos through the unmanned aerial vehicle camera, the positioning signal is a signal for providing information such as unmanned aerial vehicle positioning, speed, direction, etc., and the telemetry signal is a signal generated by a sensor carried on the unmanned aerial vehicle for monitoring information such as environment and unmanned aerial vehicle state.
In a specific embodiment, the frequency correction submodule includes a frequency compensating system and an estimating system, the frequency compensating system is used for performing fast fourier transform on a received signal to realize frequency compensation, the estimating system is used for estimating the frequency of the received signal and reducing the influence of frequency deviation on demodulation performance, and an output end of the frequency compensating system is connected with an input end of the estimating system.
In a specific embodiment, the feature extraction submodule comprises a frequency judging submodule, a local frequency extraction model, a time domain frequency extraction model and a fusion submodule, wherein the frequency judging submodule is used for judging whether the frequency of a frequency band in an initial signal is uniform or not, extracting the local frequency band with nonuniform frequency facilitates subsequent signal analysis, the local frequency extraction model adopts a local frequency basis function LFBF algorithm and an improved local three-value mode LTP algorithm, feature information of the local nonuniform frequency in an image transmission signal is effectively extracted to obtain a local frequency band static image feature, the time domain frequency extraction model adopts a time domain differential entropy TDE algorithm, time differential entropy and differential entropy of the signal are calculated in a time domain, rule information of time domain change of the signal is extracted, the fusion submodule is used for fusing the rule information of the time domain change of the nonuniform frequency band signal and the uniform frequency band signal into an information feature set, the output end of the frequency judging submodule is connected with the input end of the local frequency extraction model and the input end of the time domain frequency extraction model, and the output end of the local frequency extraction model is connected with the input end of the time domain frequency extraction model.
In a specific embodiment, the feature extraction submodule adopts a mode of dividing an initial signal into a local frequency band with uneven frequency distribution and an even frequency band with even frequency distribution, and calculating the complete feature of the local frequency band and the frequency band with even frequency distribution respectively through a time domain differential entropy TDE algorithm to obtain more comprehensive unmanned aerial vehicle signal features, and the working mode of the feature extraction submodule is as follows:
c1, extracting local frequency bands with uneven frequency distribution in an initial signal, wherein the rest frequency bands are uniform frequency bands;
calculating the complete characteristics of the local frequency band, and extracting the characteristics of the local frequency band with uneven frequency distribution in the initial signal image through the local frequency extraction model to obtain the complete characteristics of the static image of the local frequency band;
performing time sequence analysis on the complete characteristics of the local frequency band, calculating time difference and difference entropy of the characteristics of the local static image on the time domain by adopting a time domain difference entropy TDE algorithm, and extracting dynamic characteristics of the local static image;
c4, carrying out time sequence analysis on the frequency bands with uniform frequency distribution, and extracting dynamic characteristics of the uniform frequency bands by adopting time difference and difference entropy of the time domain difference entropy TDE algorithm on the uniform frequency bands in the time domain;
And C5, fusing the features, and splicing the dynamic features of the static image features and the dynamic features of the uniform frequency bands according to time sequences through the fusion sub-module to obtain a complete information feature set.
In a specific embodiment, the local frequency extraction model obtains comprehensive LTP features by combining multi-scale features of an LTP image with the LTP features, increases the accuracy of expression of the LTP features, and increases the robustness and the expressivity of spectrum features by adopting a method of converting the LTP features into LFBF features, wherein the local frequency extraction model works in the following manner:
step 1, framing a local frequency band, and dividing the local frequency band into frames with continuous time domains, wherein each second is 48 frames;
step 2, performing fast Fourier transform on each frame to obtain a frequency domain representation corresponding to the signal and the time domain;
step 3, mapping the frequency spectrum onto a logarithmic scale, and mapping the frequency spectrum into a coordinate system by adopting a Mel scale to obtain nonlinear mapping which can be perceived by human ears;
step 4, generating LTP images, and for each frame of spectrum image, calculating a three-value mode of difference between adjacent samples in a local area to generate corresponding LTP images;
step 5, confirming LTP features to obtain feature vectors of each frame, and extracting features of the LTP images by adopting an improved local three-value mode LTP algorithm to obtain the LTP features of each frame;
Step 6, confirming multi-scale characteristics of the LTP image, adopting a multi-scale filter to carry out filtering operation on the LTP image, and carrying out filtering on different scales on the LTP image of each frequency band to extract the multi-scale characteristics of the LTP image;
step 7, obtaining comprehensive LTP characteristics, and connecting the LTP characteristics obtained by adopting an improved local three-value mode LTP algorithm with the LTP characteristics obtained by adopting a multi-scale filter for filtering to obtain the comprehensive LTP characteristics;
step 8, obtaining LBF characteristics, carrying out normalization processing and binarization operation on the comprehensive LTP characteristics, wherein each basic mode is a characteristic, using the occurrence frequency to represent the importance degree of the characteristics, and superposing the occurrence frequency of each basic mode in an image to form an LBF characteristic vector;
and 9, obtaining the complete characteristics of the local frequency band, and splicing the LFBF characteristics of each frame into one characteristic which is the complete characteristics of the local frequency band.
In a specific embodiment, the improved local three-value mode LTP algorithm obtains a preliminary feature value by selecting random sampling points in a pixel neighborhood to perform pixel value comparison, describes features of a signal image more comprehensively, reduces influence of noise factors, obtains LTP features by comparing magnitude relations between the preliminary feature value of a central pixel and the preliminary feature value of a symmetrical pixel, and has better adaptability to edges of the signal image, and the working method of the improved local three-value mode LTP algorithm is as follows:
S1, confirming sampling points, selecting any pixel point in a spectrum image of a frame as a central pixel point, and selecting S random sampling points in the neighborhood of the central pixel point, wherein the expression of the number of the sampling points is as follows:
in the formula (1), S is the number of sampling points, r is the neighborhood radius, the number of pixel points in the neighborhood is x, and n is the total number of pixel points;
s2, coding the central pixel point, and calculating the size relation between the pixel value of the central pixel point and the pixel value of each sampling point, wherein the coding rule is as follows:
in the formula (2), X is the pixel value of the central pixel point, Y i For the pixel value of the sampling point, i is the index of the sampling point, Q α The method is based on a central pixel point coding value of a sampling point pixel, w is the average value of the sampling point pixel, and s is the standard deviation of the sampling point pixel;
s3, confirming a preparation characteristic value, connecting central pixel point codes based on sampling point pixels to obtain a binary code sequence with the length of 8, and converting binary into decimal to obtain the preparation characteristic value of the central pixel point;
s4, coding the symmetrical pixel points, confirming the symmetrical pixel points of the central pixel point according to a symmetry rule, and calculating the preparation characteristic values of the symmetrical pixel points according to the methods in S1, S2 and S3;
S5, recoding the central pixel point, and calculating the size relation between the central pixel point preparation characteristic value and the symmetrical pixel point preparation characteristic value, wherein the coding rule is as follows:
in the formula (3), A is a central pixel point preparation characteristic value, B is a symmetrical pixel point preparation characteristic value, Q β For central pixel point coding based on symmetrical pixels, Z j The pixel value j is the reference number of the sampling point in the adjacent symmetrical pixel point;
s6, confirming a characteristic value, namely connecting central pixel point codes based on symmetrical pixels to obtain a binary code sequence with the length of 8, and converting binary into decimal to obtain LTP characteristics of the central pixel point;
and S7, traversing pixel points in the spectrum image, and calculating LTP characteristics of all the pixel points in the spectrum image of one frame to obtain the LTP characteristics.
In a specific embodiment, the self-demodulation model includes a sample library, a classification system and a modem library, where the sample library is used to store the modulated digital signal characteristics of the marked modulation and demodulation modes, and used as a reference sample for confirming the modulation mode of the original signal, the classification system is used to confirm the modulation mode of the modulated digital signal, the modem library matches the corresponding demodulation mode for the modulated digital signal according to the modulation mode confirmed by the classification system, so as to realize demodulation of the digital signal, an output end of the sample library is connected with an input end of the classification system, and an output end of the classification system is connected with an input end of the modem library.
In a specific embodiment, the improved K-means algorithm eliminates redundant data by removing orphan points, confirms a clustering center point by a fitness sorting method, increases classification accuracy of the classification system, confirms a modulation mode of a modulated digital signal to be classified by a method of calculating a distance between a digital signal feature of a marked modulation and demodulation mode and the modulated digital signal to be classified, and the working mode of the improved K-means algorithm is as follows:
1) Confirming a K value, and fixing the K value to be 10;
2) Treating the modulated digital signal characteristic as a data point;
3) Removing the isolated points, calculating the point density of the data points, confirming that the data points with the point density lower than the density index are the isolated points, and deleting the isolated points from the data set;
4) Confirming a clustering center point, sorting data points according to the magnitude of the fitness, and selecting ten data points with the highest fitness as the clustering center point;
5) Clustering calculation, namely classifying data around a clustering center point into a data cluster to be classified according to the Euclidean distance as a radius;
6) And classifying data, namely merging the digital signal characteristics of the marked modulation and demodulation mode into the data cluster to be classified, and when the distance between the digital signal characteristics of the marked modulation and demodulation mode and the data cluster to be classified is calculated to be smaller than a distance threshold value, classifying the data cluster to be classified into the category of the digital signal characteristics of the marked modulation and demodulation mode, wherein the distance threshold value formula is as follows:
In the formula (4), P n G is the central point position of the digital signal characteristic of the marked modulation-demodulation mode m The method comprises the steps of taking a clustering center point of a data cluster to be classified, wherein k is the number of the data clusters to be classified, n is the label of the center point of the data cluster to be classified, m is the label of the clustering center point of a digital signal characteristic of a marked modulation-demodulation mode, lambda is the radius of the data cluster to be classified, v is the clustering radius of the digital signal characteristic of the marked modulation-demodulation mode, and D is a distance threshold.
Through the above embodiment, the integrity of the extracted signal features in the feature extraction submodule is shown in table 2:
table 2 integrity table of signal characteristics
According to different signal emission times, three test groups are set, and the duty ratio of a conventional task signal, a control signal, an image transmission signal, a telemetry signal and a positioning signal in the three groups of transmitted signals is 2:1:3:1:3, respectively carrying out feature extraction on three groups of signals by adopting four algorithms, comparing the feature extraction results of the four algorithms with signal sample features, calculating the signal feature integrity of the four algorithms, carrying out feature extraction on the signals by adopting a time domain differential entropy TDE algorithm, carrying out feature extraction on the signals by adopting an LTP algorithm, carrying out feature extraction on the signals by adopting a method of extracting dynamic features by adopting a time domain differential entropy TDE algorithm on the basis of the local static images, carrying out feature extraction on the signals by adopting a method of converting the LTP features into LFBF features, carrying out feature extraction on the signals by adopting a method of extracting dynamic features by adopting a time domain differential entropy TDE algorithm on the basis of the local static images, carrying out the feature extraction on the signals by adopting the algorithm of the invention, calculating the LTP features of a signal spectrum by adopting an improved local three-value mode LTP algorithm, and converting the LTP features into the LFBF features, the method for extracting dynamic characteristics by using a time domain differential entropy TDE algorithm on the basis of a local static image carries out characteristic extraction on signals, as shown in a table 2, the characteristic extraction integrity of an algorithm 2 is obviously higher than that of an algorithm 1, the method for extracting local static characteristics by using an LTP algorithm in a frequency band with signal frequency fluctuation is illustrated, the accuracy of signal characteristic extraction is improved, the characteristic extraction integrity of the algorithm 2 and the algorithm 1 is obviously reduced along with the increase of the signal receiving number when the characteristic extraction is carried out, the characteristic extraction robustness of the algorithm 3 and the characteristic extraction integrity of the algorithm 4 is obviously higher than that of the algorithm 1 and the algorithm 2 by using only the time domain differential entropy TDE algorithm or the LTP and TDE mixed algorithm, the characteristic extraction integrity of the algorithm 3 and the algorithm 4 is illustrated to be obviously higher than that of the algorithm 3, the improved local three-value mode LTP algorithm can describe the characteristics of the signal image more comprehensively, reduce the influence of noise factors, and have better adaptability to the edges of the signal image, so that the signal characteristic extraction method is a signal processing method with strong comprehensiveness, high integrity and strong robustness.
While specific embodiments of the present invention have been described above, it will be understood by those skilled in the art that these specific embodiments are by way of example only, and that various omissions, substitutions, and changes in the form and details of the methods and systems described above may be made by those skilled in the art without departing from the spirit and scope of the invention. For example, it is within the scope of the present invention to combine the above-described method steps to perform substantially the same function in substantially the same way to achieve substantially the same result. Accordingly, the scope of the invention is limited only by the following claims.
Claims (3)
1. An unmanned aerial vehicle signal spectrum identification method based on artificial intelligence is characterized in that: the method comprises the following steps:
step one, receiving unmanned aerial vehicle signals through a three-in-one antenna;
step two, storing the received original signals to a storage module;
the storage module is used for storing the received original signals sent by the unmanned aerial vehicle and used for carrying out data processing subsequently;
step three, preprocessing the signals through a data preprocessing module, and extracting signal characteristics;
the data preprocessing module carries out filtering, positive frequency and feature extraction on the signals, so that the signals are convenient to identify and analyze;
The data preprocessing module comprises a filtering submodule, a frequency correction submodule and a characteristic extraction submodule, wherein the filtering submodule adopts a band-pass filter for filtering out-of-band noise of the discretized digital signal, the frequency correction submodule is used for carrying out frequency estimation and frequency compensation on the received signal, the characteristic extraction submodule adopts a method of combining a local frequency basis function LFBF algorithm, an improved local three-value mode LTP algorithm and a time domain differential entropy TDE algorithm to extract key characteristics in the signal for subsequent spectrum analysis, the output end of the filtering submodule is connected with the input end of the frequency correction submodule, and the output end of the frequency correction submodule is connected with the input end of the characteristic extraction submodule;
step four, confirming the signal type of the original signal through a digitizing module, and digitizing the analog signal;
the digitizing module is used for ensuring that all received original signals of the unmanned aerial vehicle are converted into digital signals;
the digitizing module comprises a type judging submodule and an A/D converter, the type judging submodule is used for judging the signal type of an original signal, the signal type comprises a digital signal and an analog signal, the analog signal is sent to the A/D converter, the A/D converter carries out time discretization and amplitude discretization on the analog signal, and the output end of the judging submodule is connected with the input end of the A/D converter;
Step five, demodulating the modulated digital signal through a demodulation module;
the demodulation module is used for converting the modulated digital signals and the analog signals into demodulated digital signals;
the demodulation module comprises a modulation judging sub-module and a self-demodulation model, wherein the modulation judging sub-module is used for judging whether a digital signal is modulated or not, the modulated digital signal is sent to the self-demodulation model, the self-demodulation model adopts a demodulation mode required by an improved K-means algorithm to confirm the signal and carries out demodulation operation on the signal, and the output end of the judging sub-module is connected with the input end of the self-demodulation model;
step six, classifying the demodulated digital signals;
the classification module classifies the demodulated digital signals according to the signal frequency band, the signal power, the initial signal modulation mode and the signal receiving and transmitting mode;
the characteristic extraction submodule comprises a frequency judging submodule, a local frequency extraction model, a time domain frequency extraction model and a fusion submodule, wherein the frequency judging submodule is used for judging whether the frequency of a frequency band in an initial signal is uniform or not, extracting the local frequency band with nonuniform frequency is convenient for subsequent signal analysis, the local frequency extraction model adopts a local frequency basis function LFBF algorithm and an improved local three-value mode LTP algorithm, characteristic information of the local nonuniform frequency in an image transmission signal is effectively extracted to obtain a local frequency band static image characteristic, the time domain frequency extraction model adopts a time domain differential entropy TDE algorithm, time differential entropy and differential entropy of the signal are calculated in the time domain, rule information of time domain change of the signal is extracted, the fusion submodule is used for fusing rule information of time domain change of the signal with the uniform frequency band into an information characteristic set, the output end of the frequency judging submodule is connected with the input end of the local frequency extraction model and the time domain frequency extraction model, the output end of the local frequency extraction model is connected with the input end of the time domain frequency extraction model, and the output end of the local frequency extraction model is connected with the input end of the fusion submodule;
The local frequency extraction model adopts a mode of combining multi-scale features of an LTP image with the LTP features to obtain comprehensive LTP features, the accuracy of the expression of the LTP features is improved, the robustness and the expressivity of spectrum features are improved by adopting a method of converting the LTP features into LFBF features, and the local frequency extraction model works in the following modes:
step 1, framing a local frequency band, and dividing the local frequency band into frames with continuous time domains, wherein each second is 48 frames;
step 2, performing fast Fourier transform on each frame to obtain a frequency domain representation corresponding to the signal and the time domain;
step 3, mapping the frequency spectrum onto a logarithmic scale, and mapping the frequency spectrum into a coordinate system by adopting a Mel scale to obtain nonlinear mapping which can be perceived by human ears;
step 4, generating LTP images, and for each frame of spectrum image, calculating a three-value mode of difference between adjacent samples in a local area to generate corresponding LTP images;
step 5, confirming LTP features to obtain feature vectors of each frame, and extracting features of the LTP images by adopting an improved local three-value mode LTP algorithm to obtain the LTP features of each frame;
step 6, confirming multi-scale characteristics of the LTP image, adopting a multi-scale filter to carry out filtering operation on the LTP image, and carrying out filtering on different scales on the LTP image of each frequency band to extract the multi-scale characteristics of the LTP image;
Step 7, obtaining comprehensive LTP characteristics, and connecting the LTP characteristics obtained by adopting an improved local three-value mode LTP algorithm with the LTP characteristics obtained by adopting a multi-scale filter for filtering to obtain the comprehensive LTP characteristics;
step 8, obtaining LBF characteristics, carrying out normalization processing and binarization operation on the comprehensive LTP characteristics, wherein each basic mode is a characteristic, using the occurrence frequency to represent the importance degree of the characteristics, and superposing the occurrence frequency of each basic mode in an image to form an LBF characteristic vector;
step 9, obtaining the complete characteristics of the local frequency band, and splicing the LFBF characteristics of each frame into a characteristic which is the complete characteristics of the local frequency band;
the improved local three-value mode LTP algorithm obtains a preparation characteristic value by selecting random sampling points in a pixel point neighborhood to conduct pixel value comparison, describes characteristics of a signal image more comprehensively, reduces influence of noise factors, obtains LTP characteristics by comparing magnitude relation between the preparation characteristic value of a central pixel point and the preparation characteristic value of a symmetrical pixel point, and has better adaptability to edges of the signal image, and the working method of the improved local three-value mode LTP algorithm comprises the following steps:
S1, confirming sampling points, selecting any pixel point in a spectrum image of a frame as a central pixel point, and selecting S random sampling points in the neighborhood of the central pixel point, wherein the expression of the number of the sampling points is as follows:
（1）
in the formula (1), S is the number of sampling points, r is the neighborhood radius, x is the number of pixel points in the neighborhood, and n is the total number of pixel points;
s2, coding the central pixel point, and calculating the size relation between the pixel value of the central pixel point and the pixel value of each sampling point, wherein the coding rule is as follows:
（2）
in the formula (2), X is the pixel value of the central pixel point,for the pixel value of the sampling point i is the index of the sampling point,/>The method is based on a central pixel point coding value of a sampling point pixel, w is the average value of the sampling point pixel, and s is the standard deviation of the sampling point pixel;
s3, confirming a preparation characteristic value, connecting central pixel point codes based on sampling point pixels to obtain a binary code sequence with the length of 8, and converting binary into decimal to obtain the preparation characteristic value of the central pixel point;
s4, coding the symmetrical pixel points, confirming the symmetrical pixel points of the central pixel point according to a symmetry rule, and calculating the preparation characteristic values of the symmetrical pixel points according to the methods in S1, S2 and S3;
S5, recoding the central pixel point, and calculating the size relation between the central pixel point preparation characteristic value and the symmetrical pixel point preparation characteristic value, wherein the coding rule is as follows:
（3）
in the formula (3), A is a central pixel point preparation characteristic value, B is a symmetrical pixel point preparation characteristic value,coding for a center pixel point based on symmetrical pixels, < >>The pixel value j is the reference number of the sampling point in the adjacent symmetrical pixel point;
s6, confirming a characteristic value, namely connecting central pixel point codes based on symmetrical pixels to obtain a binary code sequence with the length of 8, and converting binary into decimal to obtain LTP characteristics of the central pixel point;
s7, traversing pixel points in the spectrum image, and calculating LTP characteristics of all the pixel points in the spectrum image of one frame to obtain LTP characteristics;
the self-demodulation model comprises a sample library, a classification system and a modem library, wherein the sample library is used for storing the modulated digital signal characteristics of a marked modulation and demodulation mode and is used as a reference sample for confirming the modulation mode of an original signal, the classification system is used for confirming the modulation mode of a modulated digital signal, the modem library is used for matching the modulated digital signal with a corresponding demodulation mode according to the modulation mode confirmed by the classification system so as to realize demodulation of the digital signal, the output end of the sample library is connected with the input end of the classification system, and the output end of the classification system is connected with the input end of the modem library;
The improved K-means algorithm eliminates redundant data by removing orphan points, confirms a clustering center point by a fitness sorting method, increases classification accuracy of the classification system, confirms a modulation mode of a modulated digital signal to be classified by a method for calculating a distance between a digital signal characteristic of a marked modulation and demodulation mode and the modulated digital signal to be classified, and works as follows:
1) Confirming a K value, and fixing the K value to be 10;
2) Treating the modulated digital signal characteristic as a data point;
3) Removing the isolated points, calculating the point density of the data points, confirming that the data points with the point density lower than the density index are the isolated points, and deleting the isolated points from the data set;
4) Confirming a clustering center point, sorting data points according to the magnitude of the fitness, and selecting ten data points with the highest fitness as the clustering center point;
5) Clustering calculation, namely classifying data around a clustering center point into a data cluster to be classified according to the Euclidean distance as a radius;
6) And classifying data, namely merging the digital signal characteristics of the marked modulation and demodulation mode into the data cluster to be classified, and when the distance between the digital signal characteristics of the marked modulation and demodulation mode and the data cluster to be classified is calculated to be smaller than a distance threshold value, classifying the data cluster to be classified into the category of the digital signal characteristics of the marked modulation and demodulation mode, wherein the distance threshold value formula is as follows:
（4）
In the formula (4), the amino acid sequence of the compound,for the central point position of the digital signal characteristic of the marked modulation-demodulation mode, < >>For the cluster center point of the data cluster to be classified, < +.>For the number of the data clusters to be classified, t is the label of the central point of the data cluster to be classified, m is the label of the clustering central point marked with the digital signal characteristics of the modulation-demodulation mode, < + >>Method for identifying the frequency spectrum of the radius numbers of a data cluster to be classified, < >>The clustering radius of the digital signal features of the marked modulation and demodulation mode is D, and the distance threshold value is D.
2. The unmanned aerial vehicle signal spectrum recognition method based on artificial intelligence according to claim 1, wherein: the frequency correction submodule comprises a frequency compensation system and an estimation system, wherein the frequency compensation system is used for carrying out fast Fourier transformation on a received signal to realize frequency compensation, the estimation system is used for estimating the frequency of the received signal and reducing the influence of frequency deviation on demodulation performance, and the output end of the frequency compensation system is connected with the input end of the estimation system.
3. The unmanned aerial vehicle signal spectrum recognition method based on artificial intelligence according to claim 1, wherein: the characteristic extraction submodule adopts a mode of dividing an initial signal into a local frequency band with uneven frequency distribution and an even frequency band with even frequency distribution, and calculating the complete characteristic of the local frequency band and the frequency band with even frequency distribution respectively through a time domain differential entropy TDE algorithm to obtain more comprehensive unmanned aerial vehicle signal characteristics, and the working mode of the characteristic extraction submodule is as follows:
C1, extracting local frequency bands with uneven frequency distribution in an initial signal, wherein the rest frequency bands are uniform frequency bands;
calculating the complete characteristics of the local frequency band, and extracting the characteristics of the local frequency band with uneven frequency distribution in the initial signal image through the local frequency extraction model to obtain the complete characteristics of the static image of the local frequency band;
performing time sequence analysis on the complete characteristics of the local frequency band, calculating time difference and difference entropy of the characteristics of the local static image on the time domain by adopting a time domain difference entropy TDE algorithm, and extracting dynamic characteristics of the local static image;
c4, carrying out time sequence analysis on the frequency bands with uniform frequency distribution, and extracting dynamic characteristics of the uniform frequency bands by adopting time difference and difference entropy of the time domain difference entropy TDE algorithm on the uniform frequency bands in the time domain;
and C5, fusing the features, and splicing the dynamic features of the static image features and the dynamic features of the uniform frequency bands according to time sequences through the fusion sub-module to obtain a complete information feature set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310691995.XA CN116582195B (en) | 2023-06-12 | 2023-06-12 | Unmanned aerial vehicle signal spectrum identification method based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310691995.XA CN116582195B (en) | 2023-06-12 | 2023-06-12 | Unmanned aerial vehicle signal spectrum identification method based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116582195A CN116582195A (en) | 2023-08-11 |
| CN116582195B true CN116582195B (en) | 2023-12-26 |
Family
ID=87537894
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310691995.XA Active CN116582195B (en) | 2023-06-12 | 2023-06-12 | Unmanned aerial vehicle signal spectrum identification method based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116582195B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106911603A (en) * | 2017-03-07 | 2017-06-30 | 北京工业大学 | A kind of broadband monitoring pattern Imitating signal modulation style real-time identification method |
| WO2018081929A1 (en) * | 2016-11-01 | 2018-05-11 | 深圳大学 | Hyperspectral remote sensing image feature extraction and classification method and system thereof |
| CN112183225A (en) * | 2020-09-07 | 2021-01-05 | 中国海洋大学 | Underwater target signal feature extraction method based on probability latent semantic analysis |
| CN112257521A (en) * | 2020-09-30 | 2021-01-22 | 中国人民解放军军事科学院国防科技创新研究院 | CNN underwater acoustic signal target identification method based on data enhancement and time-frequency separation |
| WO2021129569A1 (en) * | 2019-12-25 | 2021-07-01 | 神思电子技术股份有限公司 | Human action recognition method |
| WO2022016884A1 (en) * | 2020-07-22 | 2022-01-27 | 江苏科技大学 | Method for extracting sea surface wind speed on basis of k-means clustering algorithm |
| CN115378777A (en) * | 2022-08-25 | 2022-11-22 | 杭州电子科技大学 | Method for identifying underwater communication signal modulation mode in alpha stable distribution noise environment |
| CN115664905A (en) * | 2022-10-18 | 2023-01-31 | 东南大学 | Wi-Fi equipment identification system and method based on multi-domain physical layer fingerprint characteristics |
-
2023
- 2023-06-12 CN CN202310691995.XA patent/CN116582195B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018081929A1 (en) * | 2016-11-01 | 2018-05-11 | 深圳大学 | Hyperspectral remote sensing image feature extraction and classification method and system thereof |
| CN106911603A (en) * | 2017-03-07 | 2017-06-30 | 北京工业大学 | A kind of broadband monitoring pattern Imitating signal modulation style real-time identification method |
| WO2021129569A1 (en) * | 2019-12-25 | 2021-07-01 | 神思电子技术股份有限公司 | Human action recognition method |
| WO2022016884A1 (en) * | 2020-07-22 | 2022-01-27 | 江苏科技大学 | Method for extracting sea surface wind speed on basis of k-means clustering algorithm |
| CN112183225A (en) * | 2020-09-07 | 2021-01-05 | 中国海洋大学 | Underwater target signal feature extraction method based on probability latent semantic analysis |
| CN112257521A (en) * | 2020-09-30 | 2021-01-22 | 中国人民解放军军事科学院国防科技创新研究院 | CNN underwater acoustic signal target identification method based on data enhancement and time-frequency separation |
| CN115378777A (en) * | 2022-08-25 | 2022-11-22 | 杭州电子科技大学 | Method for identifying underwater communication signal modulation mode in alpha stable distribution noise environment |
| CN115664905A (en) * | 2022-10-18 | 2023-01-31 | 东南大学 | Wi-Fi equipment identification system and method based on multi-domain physical layer fingerprint characteristics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116582195A (en) | 2023-08-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10776659B2 (en) | Systems and methods for compressing data | |
| US9159137B2 (en) | Probabilistic neural network based moving object detection method and an apparatus using the same | |
| CN111008961A (en) | Transmission line equipment defect detection method and system, equipment and medium thereof | |
| WO2021205424A3 (en) | System and method of feature detection in satellite images using neural networks | |
| CN112395972B (en) | Unmanned aerial vehicle image processing-based insulator string identification method for power system | |
| CN114612987A (en) | Expression recognition method and device | |
| CN112632855A (en) | Method and system for detecting optical performance of multi-task convolutional network based on feature fusion | |
| Tragulnuch et al. | High voltage transmission tower detection and tracking in aerial video sequence using object-based image classification | |
| CN113723344A (en) | Video identification method and device, readable medium and electronic equipment | |
| Satrasupalli et al. | Single Image Haze Removal Based on transmission map estimation using Encoder-Decoder based deep learning architecture | |
| CN115083006A (en) | Iris recognition model training method, iris recognition method and iris recognition device | |
| CN116582195B (en) | Unmanned aerial vehicle signal spectrum identification method based on artificial intelligence | |
| CN116016071A (en) | Modulation signal identification method based on double-flow fusion CNN-BiLSTM network | |
| El-Yamany et al. | A generic approach CNN-based camera identification for manipulated images | |
| CN108304805A (en) | A kind of big data image recognition processing system | |
| CN114220003A (en) | Multi-target unsupervised domain self-adaption method for large-range ground object segmentation | |
| CN115496088A (en) | Hyperspectral image transmission method and device | |
| CN112232430A (en) | Neural network model testing method and device, storage medium and electronic device | |
| KR20210050684A (en) | Image processing system | |
| CN110555342A (en) | image identification method and device and image equipment | |
| CN110490060B (en) | Security protection front-end video equipment based on machine learning hardware architecture | |
| KR102595910B1 (en) | Vehicle number automatic recognition system and method therefor | |
| Tragulnuch et al. | High voltage transmission tower identification in an aerial video sequence using object-based image classification with geometry information | |
| Prakash et al. | Intelligent systems for redundancy removal with proficient run-length coding and statistical analysis using regression | |
| CN111476353B (en) | Super-resolution method of GAN image introducing significance |
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 |