CN114520758A - Signal modulation identification method based on instantaneous characteristics - Google Patents

Signal modulation identification method based on instantaneous characteristics Download PDF

Info

Publication number
CN114520758A
CN114520758A CN202111506378.5A CN202111506378A CN114520758A CN 114520758 A CN114520758 A CN 114520758A CN 202111506378 A CN202111506378 A CN 202111506378A CN 114520758 A CN114520758 A CN 114520758A
Authority
CN
China
Prior art keywords
instantaneous
signal
sequence
frequency
phase
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111506378.5A
Other languages
Chinese (zh)
Inventor
孙晓东
孙思瑶
刘禹震
于晓辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jilin University
Original Assignee
Jilin University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jilin University filed Critical Jilin University
Priority to CN202111506378.5A priority Critical patent/CN114520758A/en
Publication of CN114520758A publication Critical patent/CN114520758A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L27/00—Modulated-carrier systems
    • H04L27/0012—Modulated-carrier systems arrangements for identifying the type of modulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Digital Transmission Methods That Use Modulated Carrier Waves (AREA)

Abstract

A signal modulation identification method based on instantaneous characteristics belongs to the technical field of communication. The invention adopts an instantaneous power spectrum to calculate instantaneous amplitude and an instantaneous frequency sequence of a signal, uses classical Hilbert transform to calculate a phase sequence, and adopts a three-branch parallel convolutional neural network-long-time memory network to modulate and identify six communication signals according to the characteristics of amplitude shift keying signals (MASK), frequency shift keying signals (MFSK) and phase shift keying signals (MPSK), so that a step of extracting characteristic parameters after preprocessing, such as a machine learning classifier, a decision tree, a support vector machine and the like adopted by classical instantaneous statistical characteristic parameters, is not needed, the complexity of calculating the instantaneous statistical parameters is reduced, the problem of identification accuracy under low signal-to-noise ratio is improved, and the accuracy of classification and identification by single characteristics can be improved by designing a three-branch parallel fusion network structure.

Description

Signal modulation identification method based on instantaneous characteristics
Technical Field
The invention belongs to the technical field of communication, and particularly relates to a signal modulation identification method based on instantaneous characteristics.
Background
Signal modulation identification techniques are an important problem to be solved in the field of modern cooperative communication as well as in the field of non-cooperative communication. With the rapid development of communication technology in recent years, the new generation of information communication technology represented by 5G is deeply integrated with many fields, and the demand for communication quality and technology is increasing. In the military field, modulation identification is a precondition for performing actions such as interference, reconnaissance and monitoring on target communication, and plays a crucial role in war situations. In the civil field, identification and monitoring of communication signals and detection and management of frequency spectrums all require identification of modulation modes and other parameters of the signals to distinguish different users, so as to detect existence of illegal malicious users. Therefore, modulation identification technology is an indispensable technical foundation in both military and civil fields.
In the process of a modulation recognition system, the pattern recognition method based on feature extraction is widely applied to actual engineering due to the advantages of simple structure, less required prior information, less required data volume, strong adaptability and the like. As early as the nineties of the 20 th century, a plurality of instantaneous statistical characteristic parameters proposed by Nandi and Azzouz et al are called classic characteristic parameters due to the advantages of simple principle, easy extraction, stable performance and the like, and are always cited or improved by later people. Most of the existing methods based on the transient characteristics are based on the maximum value gamma max of the spectral density of zero-center normalized transient amplitude; the standard deviation sigma ap of the absolute value of the instantaneous phase nonlinear component of the zero-center non-weak signal segment; the standard deviation sigma dp of the instantaneous phase nonlinear component of the zero-center non-weak signal segment; and the standard deviation sigma aa of the zero-center normalized instantaneous amplitude absolute value and the standard deviation sigma af of the zero-center normalized non-weak signal segment instantaneous frequency absolute value are carried out by five instantaneous statistical characteristic parameters. However, the instantaneous statistical characteristic parameters are greatly influenced by noise, and the identification accuracy is very low under the condition of low signal-to-noise ratio. Therefore, under the condition of low signal-to-noise ratio, how to improve the transient characteristic anti-noise performance and improve the accuracy of final identification is an important problem to be solved at present.
The modulation identification of the communication signal mainly means that the information of the baseband signal is modulated on the amplitude, the phase or the frequency of a carrier wave, and the difference of the modulation signal on the amplitude, the phase and the frequency is the key for distinguishing different modulation modes, so the premise and the basis for identifying the modulation mode of the communication signal by using instantaneous characteristics are the accurate extraction of the instantaneous information of the modulation signal.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the communication signal modulation mode identification method based on the instantaneous characteristics is provided to solve the problem that the noise interference of the communication signal affects the identification accuracy under the condition of low signal-to-noise ratio.
A signal modulation identification method based on transient characteristics comprises the following steps which are sequentially carried out,
step one, communication signal acquisition
Respectively processing different modulation modes on a baseband signal to obtain a modulated multi-system amplitude keying signal MASK, a multi-system frequency shift keying signal MFSK or a multi-system phase shift keying signal MPSK, and performing simulation on the three types of modulation signals by using matlab software to obtain a communication modulation signal S (t) with an expression as follows:
Figure BDA0003403278640000021
in the formula: s (t) is a communication signal; t is a time variable; f. ofcIs the carrier frequency; a (t) is the instantaneous amplitude;
Figure BDA0003403278640000022
is a non-linear phase;
step two, making a signal instantaneous power spectrum
Performing an instantaneous power spectrum conversion IPS on the communication signal S (t), and obtaining a signal instantaneous power spectrum expression as follows:
Figure BDA0003403278640000023
in the formula: IPS (t, f) is the Instantaneous Power Spectrum (IPS) transform of signal s (t); t is a time variable; f is a frequency variable; τ is a time delay; s*Is the conjugation of S; h (τ) is a window function, satisfying h (0) ═ 1, where h (0) represents the value of the window function h (τ) at 0;
step three, extracting instantaneous frequency and instantaneous amplitude
Deriving, from the signal instantaneous power spectrum expression, an expression in the time domain and in the frequency domain derived from the instantaneous power spectrum properties:
Figure BDA0003403278640000031
Figure BDA0003403278640000032
wherein, s (t) represents a communication signal, and s (f) represents a fourier transform of s (t);
instantaneous frequency fi(t) the expression derived from the above equation is:
Figure BDA0003403278640000033
instantaneous frequency fi(t) the energy density at (t) represents the amplitude of the communication signal, then the instantaneous amplitude is obtained from the intensity of the energy value at the instantaneous frequency;
step four, instantaneous phase extraction
Selecting a nonlinear phase as a phase characteristic sequence, and subjecting a communication signal S (t) to Hilbert transform in a time domain to obtain an analytic signal Z (t) ═ H [ S (t)](x (t) + j) y (t), and the obtained phase is determined
Figure BDA0003403278640000034
And performing an unwrapping operation to obtain an unwrapped phase of
Figure BDA0003403278640000035
And then removing linear phase components caused by carrier frequency to obtain a nonlinear phase sequence:
Figure BDA0003403278640000036
in the formula, Z (t) is an analytic signal after Hilbert transform; x (t) is the real part of the analytic signal; y (t) is the imaginary part of the analytic signal; phi (i) is the nonlinear phase; f. ofcRepresents a carrier frequency; t is a time variable; f. ofsRepresents the sampling frequency; i represents a sampling point;
step five, sampling and dimensionality reduction are carried out on the characteristic sequence
Performing multiple dimension reduction operations of 1 sampling at intervals on the instantaneous amplitude sequence and the instantaneous frequency sequence of the signal with the length of N obtained in the third step and the instantaneous phase sequence with the length of N obtained in the fourth step, so that the long sequence with the length of N is reduced to 2000 sampling points, wherein N is a natural number;
step six, making the extracted instantaneous characteristic sequence into a data set
Selecting a multi-system amplitude keying signal MASK, a multi-system frequency shift keying signal MFSK and a multi-system phase shift keying signal MPSK, and randomly selecting M modulation signals, wherein M is a natural number, repeating the steps from one to five aiming at each communication signal, respectively obtaining M corresponding instantaneous amplitude, instantaneous frequency and instantaneous phase sequence after dimension reduction, storing the M instantaneous amplitude sequences in a csv file format to form an instantaneous amplitude sequence set, storing the M instantaneous frequency sequences in the csv file format to form an instantaneous frequency sequence set, storing the M instantaneous phase sequences in the csv file format to form an instantaneous phase sequence set, and storing the M instantaneous phase sequences in the csv file format to form the instantaneous phase sequence set, wherein the sequence data corresponding to the same sequence number in each data set come from the same signal;
seventhly, dividing the data set
Dividing three instantaneous feature sequence sets, namely an instantaneous amplitude sequence set, an instantaneous frequency sequence set and an instantaneous phase sequence set, into a training set and a verification set according to a set proportion;
step eight, constructing a convolutional neural network-long-and-short-term memory network CNN-LSTM model
Constructing a convolutional neural network-long-short-term memory network (CNN-LSTM) model fused with three branches in parallel, and synchronously training and verifying the amplitude characteristic sequence, the frequency characteristic sequence and the phase characteristic sequence corresponding to the same sequence number by utilizing the data in a training set and a verification set, so as to obtain an optimal convolutional neural network-long-short-term memory network (CNN-LSTM) model;
step nine, using a convolutional neural network-long-and-short-term memory network CNN-LSTM model to perform modulation class classification and identification
Carrying out steps one to five on the multilevel amplitude keying signal MASK, the multilevel frequency shift keying signal MFSK or the multilevel phase shift keying signal MPSK which are classified and identified in advance, and storing an instantaneous amplitude sequence, an instantaneous frequency sequence and an instantaneous phase sequence which are subjected to corresponding dimensionality reduction in a reserved corresponding sequence set as a test set in a csv file format;
and (4) inputting the CSV file which is classified and identified in advance into the optimal convolutional neural network-long-time memory network CNN-LSTM model obtained in the step eight, and identifying the modulation type of the communication signal.
The method for constructing the convolutional neural network-long-and-short-term memory network CNN-LSTM model with three parallel branches in the step eight comprises the following steps:
firstly, selecting a one-dimensional convolutional neural network for processing one-dimensional data and a long-time memory network special for processing a time sequence for each of the three branches, namely selecting a multilayer convolutional neural network-long-time memory network CNN-LSTM serial network structure;
the third branch adopts the same network structure and is arranged in sequence,
the number of one-dimensional convolutional layers, the number of pooling layers and the number of long-time and short-time memory networks (LSTM) are defined by each branch, and then the number and the size of convolutional cores in each convolutional layer, the parameters of the pooling layers and the parameters of the long-time and short-time memory layers are set;
connecting the three set branches by using a concatenate connection function, and splicing the extracted characteristics of the three network branches;
setting the number of all-connected layers and parameters of each layer in a convolutional neural network-long-term memory network CNN-LSTM and inputting the results after the characteristics are spliced in the step three;
fifthly, outputting the multi-classification task neural network by using a softmax function in an output layer in the convolutional neural network-long-time memory network CNN-LSTM, wherein the modulation type of the output communication signal is a classification recognition result.
In the step eight, the three branches input the amplitude characteristic sequence corresponding to the same sequence number one to one by using the data in the training set and the verification set, and the specific steps of synchronously training and verifying the frequency characteristic sequence and the phase characteristic sequence are as follows:
setting initial parameters of a convolutional neural network-long-short-term memory network CNN-LSTM with three branches fused in parallel, and simultaneously performing neural network training by utilizing a training set and a verification set, wherein the initial parameters comprise iteration times epoch, a learning rate, a gradient descent function and a loss function;
the convolutional neural network-long-short time memory network CNN-LSTM generates two loss value curves corresponding to the training set and the verification set after the training is finished, the two loss value curves are fitted, if under-fitting or over-fitting occurs, the step III is skipped, and if under-fitting or over-fitting does not occur, the step V is skipped;
resetting the number and parameters of the convolution layer, the pooling layer and the long and short term memory layer of each branch, and resetting the iteration times epoch, the learning rate, the gradient descent function and the loss function of the convolution neural network-long term memory network CNN-LSTM;
fourthly, aiming at the convolutional neural network-long-short-term memory network CNN-LSTM with new parameters set in the third step, repeating the first step and the second step by utilizing a training set and a verification set until under-fitting and over-fitting conditions do not occur after two loss value curves corresponding to the training set and the verification set are fitted;
and fifthly, storing the convolutional neural network-long-term memory network CNN-LSTM as an optimal convolutional neural network-long-term memory network CNN-LSTM model.
Through the design scheme, the invention can bring the following beneficial effects:
1. the invention relates to a communication signal modulation mode identification method based on instantaneous characteristics, which obtains the information of a signal time-frequency domain by carrying out instantaneous power spectrum transformation on the signal; the instantaneous frequency sequence and the instantaneous amplitude sequence can be obtained through the property of the instantaneous power spectrum. The instantaneous feature sequence is selected instead of the classical instantaneous statistical parameters as the preprocessing features, so that the calculation amount of instantaneous statistical parameters is reduced;
2. the communication signal modulation mode identification method based on the instantaneous characteristics is an instantaneous characteristic sequence obtained through time-frequency domain transformation, has the information and characteristics of a time domain and a frequency domain, can improve the sensitivity of a classic Hilbert transformation method for obtaining instantaneous statistical characteristic parameters to noise under the condition of low signal-to-noise ratio, and improves the identification accuracy under the condition of low signal-to-noise ratio;
3. the process of classifying and identifying by adopting the deep learning classifier can extract and classify and identify the characteristics of the data in the neural network, and does not need the step of extracting the characteristic parameters after preprocessing like a machine learning classifier adopted by the classic instantaneous statistical characteristic parameters, such as a decision tree, a support vector machine and the like. By designing a network structure with three branches fused in parallel, the classification accuracy of identification by single characteristics can be improved.
Drawings
The invention is further described with reference to the following figures and detailed description:
FIG. 1 is a block diagram of a signal modulation identification method based on transient characteristics according to the present invention;
FIG. 2 is a time domain waveform diagram of a 4ASK signal under a noise-free condition in an embodiment of a signal modulation identification method based on transient characteristics according to the present invention;
FIG. 3 is a time domain waveform diagram of a 4FSK signal under a noise-free condition in an embodiment of a signal modulation identification method based on temporal characteristics according to the present invention;
FIG. 4 is a time domain waveform diagram of a 4PSK signal under a noise-free condition in an embodiment of a signal modulation identification method based on transient characteristics according to the present invention;
FIG. 5 is a graph of instantaneous amplitude of a 4ASK signal under a noise-free condition in an embodiment of a transient characteristic-based signal modulation identification method of the present invention;
FIG. 6 is a comparison graph of the instantaneous amplitude of the 4ASK signal in the embodiment of the signal modulation identification method based on the instantaneous characteristics of the present invention, when the signal-to-noise ratio is 5dB, and the instantaneous power spectrum method;
FIG. 7 is a graph of instantaneous frequency of a 4FSK signal under noiseless conditions in an embodiment of a method for signal modulation identification based on temporal characteristics of the present invention;
FIG. 8 is a graph showing the comparison of the instantaneous frequency of a 4FSK signal in the case of a signal-to-noise ratio of 5dB according to an embodiment of the instantaneous characteristic-based signal modulation identification method of the present invention;
FIG. 9 is a diagram illustrating instantaneous phase and signal-to-noise ratio of a 4PSK signal at 5dB under a noise-free condition according to an embodiment of a method for identifying signal modulation based on instantaneous characteristics according to the present invention;
FIG. 10 is a diagram of sampled reduced instantaneous amplitudes in an embodiment of a method for signal modulation identification based on temporal characteristics of the present invention;
FIG. 11 is a diagram of instantaneous frequency after dimension reduction for sampling in an embodiment of a signal modulation identification method based on instantaneous characteristics according to the present invention;
FIG. 12 is a diagram of the instantaneous phase after dimension reduction for sampling in an embodiment of a signal modulation identification method based on instantaneous characteristics according to the present invention;
fig. 13 is a diagram of a neural network structure designed in an embodiment of a signal modulation identification method based on temporal characteristics according to the present invention.
Detailed Description
The technical solutions of the embodiments of the present invention are clearly and completely described below with reference to the drawings in the present invention, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
For easy understanding and explanation, as shown in fig. 1 to 13, the present invention provides a signal modulation identification method based on transient characteristics, comprising the following steps:
step one, communication signal acquisition respectively carries out processing of different modulation modes on a baseband signal to obtain a modulated multilevel amplitude keying signal MASK, a multilevel frequency shift keying signal MFSK or a multilevel phase shift keying signal MPSK, each type of modulation signal can be generated by loading the baseband signal onto a carrier signal, the generation method is mature, matlab software is used for simulating the three types of modulation signals, and the expression of a communication modulation signal S (t) can be expressed as follows:
Figure BDA0003403278640000071
in the formula: s (t) is a communication signal; t is a time variable; f. ofcIs the carrier frequency; a (t) is the instantaneous amplitude;
Figure BDA0003403278640000081
is a non-linear phase. If the signal center frequency fcThe sum bandwidth Δ f satisfies the condition fcΔ f, the signal can be considered to be a narrowband signal.
Step two, making a signal instantaneous power spectrum
The communication signal is subjected to instantaneous power spectrum transformation (IPS), which is a time-frequency analysis method proposed by Page and is rather insensitive to window parameters. Even at the analysis end point, the method maintains good time-frequency localization characteristic and has the expression of
Figure BDA0003403278640000082
In the formula: IPS (t, f) is the Instantaneous Power Spectrum (IPS) transform of signal s (t); t is a time variable; f is a frequency variable; τ is a time delay; s. the*Is the conjugation of S; h (τ) is a window function, and h (0) ═ 1 is satisfied, where h (0) denotes a value of the window function h (τ) at 0.
Step three, extracting instantaneous frequency and instantaneous amplitude
For the instantaneous power spectrum transform (IPS) expression of the communication signal s (t), the expressions in the time and frequency domain derived from the instantaneous power spectrum properties can be derived:
Figure BDA0003403278640000083
Figure BDA0003403278640000084
wherein s (t) represents a communication signal, s (f) represents a fourier transform of s (t);
instantaneous frequency fi(t) the expression derived from the above equation is:
Figure BDA0003403278640000085
instantaneous frequency fi(t) the energy density at (t) represents the amplitude of the communication signal, then the instantaneous amplitude is obtained from the intensity of the energy value at the instantaneous frequency;
step four, instantaneous phase extraction
For the extraction of the instantaneous phase sequence, the nonlinear phase is selected as the phase signature sequence. Selecting a nonlinear phase as a phase characteristic sequence, and subjecting a communication signal S (t) to Hilbert transform in a time domain to obtain an analytic signal Z (t) ═ H [ S (t)](x (t) + j) y (t), and the obtained phase is determined
Figure BDA0003403278640000091
And performing an unwrapping operation to obtain an unwrapped phase of
Figure BDA0003403278640000092
And then removing linear phase components caused by carrier frequency to obtain a nonlinear phase sequence:
Figure BDA0003403278640000093
in the formula, Z (t) is an analytic signal after Hilbert transform; x (t) is the real part of the analytic signal; y (t) is the imaginary part of the analytic signal; phi (i) is the nonlinear phase; f. ofcRepresents a carrier frequency; t is a time variable; f. ofsRepresents the sampling frequency; i represents a sampling point;
step five, sampling and dimensionality reduction are carried out on the characteristic sequence
And (3) carrying out multiple dimension reduction operations of 1-in-1-out on the signal instantaneous amplitude sequence and the instantaneous frequency sequence with the length of N obtained in the third step and the instantaneous phase sequence with the length of N obtained in the fourth step, so that the long sequence with the length of N is reduced to 2000 sampling points, wherein N is a natural number, thus not only keeping the instantaneous characteristics of the communication signal from losing, but also reducing the data length and reducing the time for subsequent neural network modulation identification.
Step six, making the extracted instantaneous characteristic sequence into a data set
Selecting communication signals of three different modulation formats of a multi-system amplitude keying signal MASK, a multi-system frequency shift keying signal MFSK and a multi-system phase shift keying signal MPSK, randomly selecting M communication signals, wherein M is a natural number and is more than or equal to 500, performing simulation generation on instantaneous amplitude, instantaneous frequency and phase sequences obtained in the mode by using matlab software, randomly generating M signals for each signal, respectively storing the obtained instantaneous amplitude, instantaneous frequency and instantaneous phase sequences into a csv file format, requiring that three data sets respectively store the instantaneous amplitude sequences, the instantaneous frequency sequences and the instantaneous phase sequences, and enabling the sequence data corresponding to each data set to be from the same signal.
Seventhly, dividing the data set
Dividing the instantaneous amplitude sequence set, the instantaneous frequency sequence set and the instantaneous phase sequence set into a training set and a verification set according to proportions, wherein data in the training set and the verification set are sequence sets of known modulation types;
step eight, constructing a convolutional neural network-long-and-short-term memory network CNN-LSTM model
Advanced techniques such as deep learning as classifiers can improve the accuracy of automatic modulation classification because they can efficiently represent features so that features are extracted and classified at the same stage. The method reduces the part for extracting the characteristic parameters, and directly inputs the data after preprocessing into the neural network for the processes of characteristic extraction and classification identification. The construction of a neural network classifier is carried out by using python software, because the frequency of the MASK signal is invariable, the amplitude obtains the maximum value at the carrier frequency, and the amplitude of the maximum value changes along with the change of the signal amplitude; the amplitude of the maximum of the MFSK signal is constant, the maximum occurring at the instantaneous frequency of the signal; MPSK signals are distinguished by the phase of the modulated signal. Six communication signals (2ASK,4ASK,2FSK,4FSK,2PSK,4PSK) cannot be identified by amplitude or frequency or phase alone;
in order to overcome the reasons, a neural network with three branches fused in parallel is constructed, the three branches respectively correspond to an amplitude characteristic sequence, a frequency characteristic sequence and a phase characteristic sequence which are input into a training set or a verification set and correspond to the same serial number one by one for synchronous training and verification,
the CSV file is used as input, training parameters are set, the set convolutional neural network-long-term memory network CNN-LSTM model is trained, and the optimal convolutional neural network-long-term memory network CNN-LSTM model is obtained and stored;
step nine, carrying out classification and identification on modulation categories of communication signals by applying convolutional neural network-long-and-short-term memory network (CNN-LSTM) model
Storing the trained neural network, storing the trained and optimally-tuned neural network model, finally performing steps one to five on a multi-system amplitude keying signal MASK, a multi-system frequency shift keying signal MFSK or a multi-system phase shift keying signal MPSK which is pre-classified and identified, and storing an instantaneous amplitude sequence, an instantaneous frequency sequence and an instantaneous phase sequence which are obtained after corresponding dimensionality reduction in a csv file format in a reserved corresponding sequence set as a test set; and (4) inputting the CSV file serving as input into the optimal convolutional neural network-long-time memory network CNN-LSTM model stored in the step eight for classification and identification, and identifying the modulation type of the communication signal.
The method for constructing the three-branch parallel-fused convolutional neural network-long-and-short-term memory network CNN-LSTM model specifically comprises the following steps:
firstly, selecting a one-dimensional convolutional neural network for processing one-dimensional data and a long-time memory network specially for processing a time sequence for each branch in the three branches, namely selecting a serial network structure of a multilayer convolutional neural network-the long-time memory network (CNN-LSTM);
the third branch adopts the same network structure and is arranged in sequence,
the number of dimension convolution layers of each branch is defined as I, the number of pooling layers is J, the number of long-time memory networks LSTM is K, and then the number and the size of convolution cores in each convolution layer, the parameters of the pooling layers and the parameters of the long-time memory layers are set;
connecting the three set branches by using a concatenate connection function, and splicing the extracted characteristics of the three network branches;
setting the number of all-connected layers and parameters of each layer in a convolutional neural network-long-term memory network CNN-LSTM and inputting the results of the splicing characteristics in the step three;
fifthly, outputting the multi-classification task neural network by using a softmax function in an output layer of the convolutional neural network-long-time memory network CNN-LSTM, wherein the modulation type of the output communication signal is the classification recognition result.
In the step eight, the three branches input the amplitude characteristic sequence corresponding to the same sequence number one to one by using the data in the training set and the verification set, and the specific steps of synchronously training and verifying the frequency characteristic sequence and the phase characteristic sequence are as follows:
setting initial parameters of a convolutional neural network-long-time memory network CNN-LSTM with three branches fused in parallel, and simultaneously carrying out neural network training by utilizing a training set and a verification set, wherein the initial parameters comprise iteration times epoch, a learning rate, a gradient descent function and a loss function;
and secondly, the convolutional neural network-the long-term memory network CNN-LSTM can generate two accurate value curves corresponding to the training set and the verification set and two loss value curves corresponding to the training set and the verification set after the training is finished. With the increase of the accurate value, the loss value correspondingly decreases, and due to the corresponding relation between the loss value and the accurate value, the training condition is judged by only utilizing the loss value curve. The loss value curve gradually flattens as the number of iterations increases until the loss value curve decreases later. When the loss value curve is flat, the training of the convolutional neural network-long-and-short-term memory network CNN-LSTM is basically completed, but the training is not necessarily optimal, and the fitting degree of the two loss value curves is also shown. Fitting the two loss value curves corresponding to the training set and the verification set, and skipping to the third step if under-fitting or over-fitting occurs, or skipping to the fifth step if under-fitting or over-fitting does not occur;
resetting the number and parameters of the convolution layer, the pooling layer and the long and short term memory layer of each branch, and resetting the parameters of the convolution neural network-the long and short term memory network CNN-LSTM such as iteration times epoch, learning rate, gradient descent function and loss function;
fourthly, aiming at the convolutional neural network-long-short-term memory network CNN-LSTM with new parameters set in the third step, repeating the first step and the second step by utilizing a training set and a verification set until under-fitting and over-fitting conditions do not occur after two loss value curves corresponding to the training set and the verification set are fitted;
and fifthly, storing the convolutional neural network-long-term memory network CNN-LSTM as an optimal convolutional neural network-long-term memory network CNN-LSTM model.
Example (b):
in order to make the method described in this embodiment better understood by those skilled in the art, the method is described below by referring to a specific example.
Simulation conditions are as follows: the carrier frequency is 1000Hz, the sampling frequency is 40000Hz, the symbol rate is 50B, the number of code elements is 16, the number of sampling points is 12800, a signal sequence S (n) is obtained, n is 1, 2, the.
1. Communication signal acquisition
Setting modulation parameters and carrying out simulation acquisition on the communication signals MASK, MFSK and MPSK, wherein each type of modulation signals can be generated by loading baseband signals onto carrier signals, and the generation method tends to mature
The expression of the communication modulation signal S (t) can be expressed as
Figure BDA0003403278640000121
In the formula: s (t) is a communication signal; t is a time variable; f. ofcIs the carrier frequency; a (t) is the instantaneous amplitude;
Figure BDA0003403278640000122
is a non-linear phase. If the signal center frequency fcThe sum bandwidth deltaf satisfies the condition fcΔ f, the signal can be considered a narrowband signal.
And (3) simulating by using matlab software, and acquiring a communication signal modulation result under the simulation condition. Fig. 2 shows a time-domain waveform of a 4ASK signal under a noise-free condition, fig. 3 shows a time-domain waveform of a 4FSK signal under a noise-free condition, and fig. 4 shows a time-domain waveform of a 4PSK signal under a noise-free condition.
2. Instantaneous power spectrum of the signal
The communication signal s (t) is subjected to an instantaneous power spectrum transformation (IPS), which is a time-frequency analysis method proposed by Page and is rather insensitive to window parameters. Even at the analysis end point, the method maintains good time-frequency localization characteristic and has the expression of
Figure BDA0003403278640000131
In the formula: IPS (t, f) is an Instantaneous Power Spectrum (IPS) transform of signal s (t); t is a time variable; f is a frequency variable; τ is a time delay; s*Is the conjugation of S; h (τ) is a window function, satisfying h (0) ═ 1, where h (0) represents the value of the window function h (τ) at 0;
3. instantaneous frequency and amplitude extraction
For the instantaneous power spectrum transform (IPS) expression of the communication signal s (t), the expressions in the time and frequency domain derived from the instantaneous power spectrum properties can be derived:
Figure BDA0003403278640000132
Figure BDA0003403278640000133
wherein, s (t) represents a communication signal, and s (f) represents a fourier transform of s (t);
instantaneous frequency fi(t) the expression derived from the above equation is:
Figure BDA0003403278640000134
instantaneous frequency fi(t) the energy density at (t) represents the amplitude of the communication signal, then the instantaneous amplitude is obtained from the intensity of the energy value at the instantaneous frequency;
because the frequency of the MASK signal is constant, the amplitude takes the maximum value at the carrier frequency, and the amplitude of the maximum value changes along with the change of the signal amplitude; the instantaneous amplitude is observed for example with a 4ASK signal. Fig. 5 is a graph showing the instantaneous amplitude of the 4ASK signal under the noise-free condition, and fig. 6 is a graph showing the instantaneous amplitude of the 4ASK signal in comparison with the instantaneous power spectrum when the signal-to-noise ratio is 5 dB;
since the amplitude of the maximum of the MFSK signal is constant, the maximum occurs at the instantaneous frequency of the signal; the instantaneous frequency is observed for example with a 4FSK signal. FIG. 7 is a graph of instantaneous frequency of a 4FSK signal under noiseless conditions, and FIG. 8 is a graph of instantaneous frequency versus the classical and instantaneous power spectrum methods for a 4FSK signal at a 5dB signal-to-noise ratio;
4. instantaneous phase extraction
For the extraction of the instantaneous phase sequence, the nonlinear phase is selected as the phase signature sequence. Selecting a nonlinear phase as a phase characteristic sequence, and subjecting a communication signal S (t) to Hilbert transform in a time domain to obtain an analytic signal Z (t) ═ H [ S (t)]X (t) + jy (t), and the obtained phase is measured
Figure BDA0003403278640000141
And performing an unwrapping operation to obtain an unwrapped phase of
Figure BDA0003403278640000142
And then removing linear phase components caused by carrier frequency to obtain a nonlinear phase sequence:
Figure BDA0003403278640000143
in the formula, Z (t) is an analytic signal after Hilbert transform; x (t) is the real part of the analytic signal; y (t) is the imaginary part of the analytic signal; phi (i) is the nonlinear phase; f. ofcRepresents a carrier frequency; t is a time variable; f. ofsRepresents the sampling frequency; i represents a sampling point;
FIG. 9 is a plot of instantaneous phase of a 4PSK signal under noise-free conditions versus the instantaneous phase of the 4PSK signal at a 5dB signal-to-noise ratio;
5. sampling and dimensionality reduction of feature sequence
The obtained instantaneous amplitude sequence, instantaneous frequency sequence and instantaneous phase sequence of the signal with the length of 12800 are subjected to 1 sampling operation for a plurality of times, so that the length of the long sequence with the length of 12800 is reduced to the length of 1600 sequences in 2000 sampling points, the instantaneous characteristics of the communication signal are kept, the data length is reduced, and the time for subsequent neural network modulation identification can be shortened.
FIG. 10 is a graph of the instantaneous amplitude after dimension reduction; FIG. 11 is a graph of instantaneous frequency after sample dimensionality reduction; FIG. 12 is a diagram of the instantaneous phase after dimension reduction;
6. extracting transient feature sequences as data sets
Respectively using matlab software to simulate and generate M multiple amplitude keying signals MASK, M multiple frequency shift keying signals MFSK and M multiple phase shift keying signals MPSK through the above mode, wherein each signal randomly generates 1000 signals, and the instantaneous amplitude, instantaneous frequency and instantaneous phase sequence obtained by each signal are respectively stored in a csv file format, which is ensured when a data set is generated, 1) three data sets are required to respectively store the instantaneous amplitude sequence, the instantaneous frequency sequence and the instantaneous phase sequence, 2) sequence data corresponding to each data set come from the same signal sample, 3) the signal sample is labeled, three characteristic sequences of each signal sample are labeled with the same label when the label is printed, and 4) each type of signal is labeled with the same label.
7. Convolutional neural network-long-short-term memory network CNN-LSTM model construction
Advanced techniques such as deep learning as classifiers can improve the accuracy of automatic modulation classification because they can efficiently represent features so that features are extracted and classified at the same stage. The method reduces the part for extracting the characteristic parameters, and directly inputs the data after preprocessing into the neural network for the processes of characteristic extraction and classification identification.
Building a neural network classifier by using python software, building a three-branch parallel fusion neural network, inputting an amplitude characteristic sequence, a frequency characteristic sequence and a phase characteristic sequence into each branch respectively, and setting the length of an input sequence of each branch to be 1600.
Setting a structure and parameters of a convolutional neural network-long-short-term memory network CNN-LSTM model, wherein the neural network selected by each branch is a multi-layer convolutional neural network-long-short-term memory network CNN-LSTM serial network structure, firstly setting 10 one-dimensional convolutional layers and 4 maximum pooling layers, setting 10 convolutional layers to adopt one-dimensional convolution, and connecting the first convolutional layer and the second convolutional layer to each other by using 64 convolutional kernels with the size of 1 × 15; the third convolutional layer and the fourth convolutional layer adopt 128 convolutional kernels with the size of 1 x 13, and then the pooling layers are connected; the fifth convolution layer to the seventh convolution layer all adopt 256 convolution kernels with the size of 1 x 11, and then the pooling layers are connected; the eighth convolutional layer to the tenth convolutional layer adopt 512 convolutional kernels with the size of 1 x 9, and then are connected with the pooling layer; the four pooling layers are all selected to be maximum pooling with the size of 1 × 2, then the parameter of the long-time memory layer is set to be 10, then the three branches are connected together through a concatenate function, then two full-connection layers are arranged, the number of neurons is 256 and 128 respectively, the number of activation functions is equal to or greater than a relu activation function, finally, the softmax activation function is selected as an output layer, and the number of neurons in the output layer is 6;
8. inputting the extracted instantaneous characteristic sequence into a convolutional neural network-long-short term memory network (CNN-LSTM) model for modulation recognition
Dividing original 1000 data sets of each signal according to the proportion of 70% of a training set and 30% of a verification set, simultaneously training a set CNN-LSTM network by utilizing the training set and the verification set, and taking a CSV file as the input of three network branches, setting training parameters and training the set convolutional neural network, wherein the three different instantaneous characteristic sequences are an instantaneous amplitude sequence, an instantaneous frequency sequence and an instantaneous phase sequence of known modulation types of training data;
adjusting super parameters such as learning rate, iteration times, loss functions and an optimizer of the neural network according to whether overfitting or under-fitting occurs on a loss function value curve drawn by the training set and the verification set, and updating internal weight parameters of the model through a back propagation algorithm until optimal network parameters are reached; storing the training and adjusting the optimal convolutional neural network-long-short-term memory network CNN-LSTM model;
9. and (3) repeating the steps 1-5, regenerating 300 signal samples, respectively obtaining 300 instantaneous amplitude sequences, instantaneous frequency sequences and instantaneous phase sequences, respectively manufacturing the instantaneous amplitude sequences, the instantaneous frequency sequences and the instantaneous phase sequences into a CSV file serving as a test set, and inputting the test set of the three instantaneous characteristic sequences into the trained optimal convolutional neural network-long-short-term memory network CNN-LSTM model for classification and identification.
The classic instantaneous statistical characteristic parameter extraction method is sensitive to noise under low signal-to-noise ratio, and meanwhile, the problem that parameter extraction needs to be carried out after signal preprocessing when classification recognition is carried out by a decision tree, a support vector machine and a BP neural network classifier which are mostly adopted in the classic method. The invention provides a communication signal modulation mode identification method based on instantaneous characteristics, which aims at the characteristics of MASK, MFSK and MPSK (specifically six communication signals of 2ASK,4ASK,2FSK,4FSK,2PSK and 4PSK), adopts instantaneous power spectrum to calculate the instantaneous amplitude and instantaneous frequency sequence of the signal, uses classical Hilbert transform to calculate the phase sequence, and simultaneously adopts a three-branch parallel convolutional neural network-long-short-term memory network to classify and identify a one-dimensional instantaneous characteristic sequence, thereby reducing the step of extracting characteristic parameters.

Claims (3)

1. A signal modulation identification method based on transient characteristics is characterized in that: comprises the following steps which are sequentially carried out,
step one, communication signal acquisition
Respectively processing different modulation modes on a baseband signal to obtain a modulated multi-system amplitude keying signal MASK, a multi-system frequency shift keying signal MFSK or a multi-system phase shift keying signal MPSK, and performing simulation on the three types of modulated signals by using matlab software to obtain a modulated communication signal S (t) with an expression as follows:
Figure FDA0003403278630000011
in the formula: s (t) is a communication signal; t is a time variable; f. ofcIs the carrier frequency; a (t) is the instantaneous amplitude;
Figure FDA0003403278630000015
is a non-linear phase;
step two, making signal instantaneous power spectrum
Performing instantaneous power spectrum conversion IPS on the communication signal S (t), and obtaining a signal instantaneous power spectrum expression as follows:
Figure FDA0003403278630000012
in the formula: IPS (t, f) is the Instantaneous Power Spectrum (IPS) transform of signal s (t); t is a time variable; f is a frequency variable; τ is a time delay; s. the*Is the conjugation of S; h (τ) is a window function, satisfying h (0) ═ 1, where h (0) represents the value of the window function h (τ) at 0;
step three, extracting instantaneous frequency and instantaneous amplitude
Deriving, from the signal instantaneous power spectrum expression, an expression in the time domain and in the frequency domain derived from the instantaneous power spectrum properties:
Figure FDA0003403278630000013
Figure FDA0003403278630000014
wherein, s (t) represents a communication signal, and s (f) represents a fourier transform of s (t);
instantaneous frequency fi(t) the expression derived from the above equation is:
Figure FDA0003403278630000021
instantaneous frequency fi(t) the energy density at (t) represents the amplitude of the communication signal, then the instantaneous amplitude is obtained from the intensity of the energy value at the instantaneous frequency;
step four, instantaneous phase extraction
Selecting a nonlinear phase as a phase characteristic sequence, and subjecting a communication signal S (t) to Hilbert transform in a time domain to obtain an analytic signal Z (t) ═ H [ S (t)](x (t) + j) y (t), and the obtained phase is determined
Figure FDA0003403278630000022
And performing an unwrapping operation to obtain an unwrapped phase of
Figure FDA0003403278630000023
And then removing linear phase components caused by carrier frequency to obtain a nonlinear phase sequence:
Figure FDA0003403278630000024
in the formula, Z (t) is an analytic signal after Hilbert transform; x (t) is the real part of the analytic signal; y (t) is the imaginary part of the analytic signal; phi (i) is the nonlinear phase; f. ofcRepresents a carrier frequency; t is a time variable; f. ofsRepresents the sampling frequency; i represents a sampling point;
step five, sampling and dimensionality reduction are carried out on the characteristic sequence
Performing multiple dimension reduction operations of 1 sampling at intervals on the instantaneous amplitude sequence and the instantaneous frequency sequence of the signal with the length of N obtained in the third step and the instantaneous phase sequence with the length of N obtained in the fourth step, so that the long sequence with the length of N is reduced to 2000 sampling points, wherein N is a natural number;
step six, making the extracted instantaneous characteristic sequence into a data set
Selecting a multi-system amplitude keying signal MASK, a multi-system frequency shift keying signal MFSK and a multi-system phase shift keying signal MPSK, and randomly selecting M modulation signals, wherein M is a natural number, repeating the steps from one to five aiming at each communication signal, respectively obtaining M corresponding instantaneous amplitude, instantaneous frequency and instantaneous phase sequence after dimension reduction, storing the M instantaneous amplitude sequences in a csv file format to form an instantaneous amplitude sequence set, storing the M instantaneous frequency sequences in the csv file format to form an instantaneous frequency sequence set, storing the M instantaneous phase sequences in the csv file format to form an instantaneous phase sequence set, and storing the M instantaneous phase sequences in the csv file format to form the instantaneous phase sequence set, wherein the sequence data corresponding to the same sequence number in each data set come from the same signal;
seventhly, dividing the data set
Dividing three instantaneous feature sequence sets, namely an instantaneous amplitude sequence set, an instantaneous frequency sequence set and an instantaneous phase sequence set, into a training set and a verification set according to a set proportion;
step eight, constructing a convolutional neural network-long-and-short-term memory network CNN-LSTM model
Constructing a convolutional neural network-long-short-term memory network (CNN-LSTM) model fused with three branches in parallel, and synchronously training and verifying the amplitude characteristic sequence, the frequency characteristic sequence and the phase characteristic sequence corresponding to the same sequence number by utilizing the data in a training set and a verification set, so as to obtain an optimal convolutional neural network-long-short-term memory network (CNN-LSTM) model;
step nine, using a convolutional neural network-long-and-short-term memory network CNN-LSTM model to perform modulation class classification and identification
Carrying out steps one to five on the multilevel amplitude keying signal MASK, the multilevel frequency shift keying signal MFSK or the multilevel phase shift keying signal MPSK which are classified and identified in advance, and storing an instantaneous amplitude sequence, an instantaneous frequency sequence and an instantaneous phase sequence which are subjected to corresponding dimensionality reduction in a reserved corresponding sequence set as a test set in a csv file format;
and (4) inputting the CSV file which is classified and identified in advance into the optimal convolutional neural network-long-time memory network CNN-LSTM model obtained in the step eight, and identifying the modulation type of the communication signal.
2. A method as claimed in claim 1, wherein the signal modulation identification method based on temporal characteristics comprises: the method for constructing the convolutional neural network-long-and-short-term memory network CNN-LSTM model with three parallel branches in the step eight comprises the following steps:
firstly, selecting a one-dimensional convolutional neural network for processing one-dimensional data and a long-time memory network special for processing a time sequence for each of the three branches, namely selecting a multilayer convolutional neural network-long-time memory network CNN-LSTM serial network structure;
the third branch adopts the same network structure and is arranged in sequence,
the number of one-dimensional convolutional layers, the number of pooling layers and the number of long-time and short-time memory networks (LSTM) are defined by each branch, and then the number and the size of convolutional cores in each convolutional layer, the parameters of the pooling layers and the parameters of the long-time and short-time memory layers are set;
connecting the three set branches by using a concatenate connection function, and splicing the extracted characteristics of the three network branches;
setting the number of all-connected layers and parameters of each layer in a convolutional neural network-long-term memory network CNN-LSTM and inputting the results after the characteristics are spliced in the step three;
fifthly, outputting the multi-classification task neural network by using a softmax function in an output layer in the convolutional neural network-long-time memory network CNN-LSTM, wherein the modulation type of the output communication signal is the classification recognition result.
3. A method as claimed in claim 2, wherein the method comprises the steps of: in the step eight, the three branches input the amplitude characteristic sequence corresponding to the same sequence number one to one by using the data in the training set and the verification set, and the specific steps of synchronously training and verifying the frequency characteristic sequence and the phase characteristic sequence are as follows:
setting initial parameters of a convolutional neural network-long-time memory network CNN-LSTM with three branches fused in parallel, and simultaneously carrying out neural network training by utilizing a training set and a verification set, wherein the initial parameters comprise iteration times epoch, a learning rate, a gradient descent function and a loss function;
the convolutional neural network-long-short time memory network CNN-LSTM generates two loss value curves corresponding to the training set and the verification set after the training is finished, the two loss value curves are fitted, if under-fitting or over-fitting occurs, the step III is skipped, and if under-fitting or over-fitting does not occur, the step V is skipped;
resetting the number and parameters of the convolution layer, the pooling layer and the long and short term memory layer of each branch, and resetting the iteration times epoch, the learning rate, the gradient descent function and the loss function of the convolution neural network-long term memory network CNN-LSTM;
fourthly, aiming at the convolutional neural network-long-short-term memory network CNN-LSTM with new parameters set in the third step, repeating the first step and the second step by utilizing a training set and a verification set until under-fitting and over-fitting conditions do not occur after two loss value curves corresponding to the training set and the verification set are fitted;
and fifthly, storing the convolutional neural network-long-short-term memory network CNN-LSTM as an optimal convolutional neural network-long-short-term memory network CNN-LSTM model.
CN202111506378.5A 2021-12-10 2021-12-10 Signal modulation identification method based on instantaneous characteristics Pending CN114520758A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111506378.5A CN114520758A (en) 2021-12-10 2021-12-10 Signal modulation identification method based on instantaneous characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111506378.5A CN114520758A (en) 2021-12-10 2021-12-10 Signal modulation identification method based on instantaneous characteristics

Publications (1)

Publication Number Publication Date
CN114520758A true CN114520758A (en) 2022-05-20

Family

ID=81596812

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111506378.5A Pending CN114520758A (en) 2021-12-10 2021-12-10 Signal modulation identification method based on instantaneous characteristics

Country Status (1)

Country Link
CN (1) CN114520758A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117807529A (en) * 2024-02-29 2024-04-02 南京工业大学 Modulation mode identification method and system for output signals of signal generator
CN118075914A (en) * 2024-04-18 2024-05-24 雅安数字经济运营有限公司 NVR and IPC automatic wireless code matching connection method
CN118432747A (en) * 2024-07-01 2024-08-02 杭州捷孚电子技术有限公司 Monitoring receiver signal control method and system

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117807529A (en) * 2024-02-29 2024-04-02 南京工业大学 Modulation mode identification method and system for output signals of signal generator
CN117807529B (en) * 2024-02-29 2024-05-07 南京工业大学 Modulation mode identification method and system for output signals of signal generator
CN118075914A (en) * 2024-04-18 2024-05-24 雅安数字经济运营有限公司 NVR and IPC automatic wireless code matching connection method
CN118432747A (en) * 2024-07-01 2024-08-02 杭州捷孚电子技术有限公司 Monitoring receiver signal control method and system

Similar Documents

Publication Publication Date Title
CN110826630B (en) Radar interference signal feature level fusion identification method based on deep convolutional neural network
Zhang et al. An efficient deep learning model for automatic modulation recognition based on parameter estimation and transformation
CN114520758A (en) Signal modulation identification method based on instantaneous characteristics
CN109890043B (en) Wireless signal noise reduction method based on generative countermeasure network
CN111428817A (en) Defense method for resisting attack by radio signal identification
CN113435247B (en) Intelligent recognition method, system and terminal for communication interference
CN113488060B (en) Voiceprint recognition method and system based on variation information bottleneck
CN110120926A (en) Modulation mode of communication signal recognition methods based on evolution BP neural network
CN116866129A (en) Wireless communication signal detection method
CN111050315B (en) Wireless transmitter identification method based on multi-core two-way network
CN112364729A (en) Modulation identification method based on characteristic parameters and BP neural network
CN111382803B (en) Communication signal feature fusion method based on deep learning
CN112115821B (en) Multi-signal intelligent modulation mode identification method based on wavelet approximate coefficient entropy
CN114726692B (en) SERESESESENet-LSTM-based radiation source modulation mode identification method
Yang et al. One-dimensional deep attention convolution network (ODACN) for signals classification
CN116797796A (en) Signal identification method based on time-frequency analysis and deep learning under DRFM intermittent sampling
CN113225282A (en) Communication signal modulation identification method based on BP neural network
CN114943245A (en) Automatic modulation recognition method and device based on data enhancement and feature embedding
CN113111786A (en) Underwater target identification method based on small sample training image convolutional network
CN113569773B (en) Interference signal identification method based on knowledge graph and Softmax regression
CN116471154A (en) Modulation signal identification method based on multi-domain mixed attention
CN114980122A (en) Small sample radio frequency fingerprint intelligent identification system and method
Zha et al. Intelligent identification technology for high‐order digital modulation signals under low signal‐to‐noise ratio conditions
CN113962260A (en) Radar signal intelligent sorting method based on denoising depth residual error network
Ma et al. DBN based automatic modulation recognition for ultra-low SNR RFID signals

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