A signal modulation mode recognition method and system based on time-frequency domain image matching

Abstract

The application provides a signal modulation mode recognition method and system based on time-frequency domain image matching, and belongs to the technical field of signal processing. The method comprises the following steps: obtaining a signal data set, which contains M types of signals with different signal system modes, and each type contains N signals; performing time-frequency analysis on the M*N signals respectively to obtain M*N time-frequency power spectrum matrices and generate M*N signal time-frequency domain images; dividing the M*N signal time-frequency domain images into a training set and a test set according to a proportion to train a recognition classifier based on a neural network; obtaining a received radio frequency signal to be recognized, performing down-conversion processing on the radio frequency signal to obtain a digital intermediate frequency signal, and generating a signal time-frequency domain image of the digital intermediate frequency signal; and using the trained recognition classifier to match the signal time-frequency domain image of the digital intermediate frequency signal with the M*N signal time-frequency domain images to obtain a recognition result of the signal modulation mode of the radio frequency signal to be recognized.

Description

Technical Field

[0001] This invention belongs to the field of signal processing technology, and in particular relates to a method and system for identifying signal modulation patterns based on time-frequency domain image matching. Background Technology

[0002] The current electromagnetic environment is highly complex, with multiple signals potentially coexisting at the same frequency, including terrestrial communication signals, satellite communication signals, navigation signals, radar signals, and so on. This mixing of signals degrades signal quality. Real-time and accurate identification of the modulation methods of different signal types can significantly improve the speed and accuracy of subsequent signal processing.

[0003] Currently, the main approach in China for signal modulation pattern recognition is to first extract the features of the signal modulation pattern and then manually set a threshold for discrimination. However, this method has low efficiency and accuracy in identifying signal modulation patterns. Summary of the Invention

[0004] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, this invention proposes a signal modulation scheme identification method based on time-frequency domain image matching, which can improve the efficiency and accuracy of signal modulation scheme identification.

[0005] The first aspect of this invention discloses a method for identifying signal modulation schemes based on time-frequency domain image matching. The method includes:

[0006] Step S1: Obtain a signal dataset. The signal dataset contains M types of signals with different signal systems. Each type contains N signals, for a total of M×N signals.

[0007] Step S2: Perform time-frequency analysis on the M×N signals in the signal dataset to obtain M×N time-frequency power spectrum matrices, and use the M×N video power spectrum matrices to generate M×N signal time-frequency domain images;

[0008] Step S3: Divide the M×N signal time-frequency domain images into a training set and a test set according to the ratio, and use the training set and test set to train and test the neural network-based recognition classifier respectively;

[0009] Step S4: Obtain the received radio frequency signal to be identified, and perform down-conversion processing on the radio frequency signal to obtain a digital intermediate frequency signal;

[0010] Step S5: Perform time-frequency analysis on the digital intermediate frequency signal to obtain the time-frequency power spectral density of the digital intermediate frequency signal, and generate a signal time-frequency domain image of the digital intermediate frequency signal;

[0011] Step S6: Using the trained recognition classifier, the recognition result of the signal modulation mode of the radio frequency signal to be identified is obtained by matching the signal time-frequency domain image of the digital intermediate frequency signal with the M×N signal time-frequency domain images.

[0012] According to the method of the first aspect of the present invention, in step S1, with a specified frequency point as the center frequency point, within a specified bandwidth range, the M×N signals are generated according to the function expression of the signal, the set sampling rate, the signal duration, and the signal size N of each type of signal to form the signal dataset.

[0013] According to the method of the first aspect of the present invention, the signals of different modulation methods include at least linear frequency modulated signals, single-frequency signals, spread spectrum signals, and pulse signals; in step S1, when generating N signals for each type of modulation method, the N signals of that type of modulation method are generated by adjusting the signal parameters, and the adjusted signal parameters specifically include:

[0014] For the linear frequency modulated signal, the adjusted signal parameters include the starting frequency, ending frequency, starting power, ending power, and frequency modulation rate;

[0015] For the single-frequency signal, the adjusted signal parameters include carrier frequency and carrier power;

[0016] For the spread spectrum signal, the adjusted signal parameters include the spread spectrum code rate, carrier frequency, and carrier power;

[0017] For the pulse signal, the adjusted signal parameters include pulse period, duty cycle, and carrier power.

[0018] According to the method of the first aspect of the present invention, the horizontal axis of the signal time-frequency domain image is time, the vertical axis is frequency, and the signal power at the coordinate point is characterized by the color or grayscale of the coordinate point; in step S2, short-time Fourier transforms are performed on the M×N signals in the signal dataset respectively to calculate the M×N time-frequency power spectrum matrices.

[0019] According to the method of the first aspect of the present invention, in step S3, training the recognition classifier specifically includes: optimizing the recognition classifier using the training set, recognizing images in the test set using the optimized recognition classifier, completing the training when the recognition classification result exceeds a threshold, otherwise continuing the training.

[0020] According to the method of the first aspect of the present invention, in step S4, a signal having the signal duration within the specified bandwidth range of the radio frequency signal is selected, and the downconversion process is performed.

[0021] According to the method of the first aspect of the present invention, in step S6, the similarity between the signal time-frequency domain image of the digital intermediate frequency signal and the M×N signal time-frequency domain images is calculated using the trained recognition classifier, and the recognition result of the signal modulation mode of the radio frequency signal to be identified is determined based on the calculation result of the similarity.

[0022] A second aspect of this invention discloses a signal modulation scheme recognition system based on time-frequency domain image matching. The system includes:

[0023] The first processing unit is configured to acquire a signal dataset, the signal dataset containing M types of signals with different signal systems, each type containing N signals, for a total of M×N signals;

[0024] The second processing unit is configured to perform time-frequency analysis on the M×N signals in the signal dataset to obtain M×N time-frequency power spectrum matrices, and use the M×N video power spectrum matrices to generate M×N signal time-frequency domain images.

[0025] The third processing unit is configured to divide the M×N signal time-frequency domain images into a training set and a test set according to the ratio, and to train and test a neural network-based recognition classifier using the training set and the test set respectively.

[0026] The fourth processing unit is configured to acquire the received radio frequency signal to be identified and perform down-conversion processing on the radio frequency signal to obtain a digital intermediate frequency signal.

[0027] The fifth processing unit is configured to perform time-frequency analysis on the digital intermediate frequency signal to obtain the time-frequency power spectral density of the digital intermediate frequency signal and generate a signal time-frequency domain image of the digital intermediate frequency signal.

[0028] The sixth processing unit is configured to use the trained recognition classifier to obtain the recognition result of the signal modulation mode of the radio frequency signal to be identified by matching the signal time-frequency domain image of the digital intermediate frequency signal with the M×N signal time-frequency domain images.

[0029] According to the system of the second aspect of the present invention, the first processing unit is specifically configured to, in step S1, generate the M×N signals within a specified bandwidth range, using a specified frequency point as the center frequency point, based on the function expression of the signal, the set sampling rate, the signal duration, and the signal size N of each type of signal, to form the signal dataset.

[0030] According to the system of the second aspect of the present invention, the signals of different signal modulation methods include at least linear frequency modulation signals, single-frequency signals, spread spectrum signals, and pulse signals; the first processing unit is specifically configured to generate N signals of each type of signal modulation method by adjusting signal parameters when generating N signals for each type of signal modulation method, wherein the adjusted signal parameters specifically include:

[0031] For the linear frequency modulated signal, the adjusted signal parameters include the starting frequency, ending frequency, starting power, ending power, and frequency modulation rate;

[0032] For the single-frequency signal, the adjusted signal parameters include carrier frequency and carrier power;

[0033] For the spread spectrum signal, the adjusted signal parameters include the spread spectrum code rate, carrier frequency, and carrier power;

[0034] For the pulse signal, the adjusted signal parameters include pulse period, duty cycle, and carrier power.

[0035] According to a system of a second aspect of the present invention, the horizontal axis of the signal time-frequency domain image is time, the vertical axis is frequency, and the signal power at the coordinate point is characterized by the color or grayscale of the coordinate point; the second processing unit is specifically configured to perform short-time Fourier transform on each of the M×N signals in the signal dataset to calculate the M×N time-frequency power spectrum matrices.

[0036] According to the system of the second aspect of the present invention, the third processing unit is specifically configured to train the recognition classifier by: optimizing the recognition classifier using the training set, recognizing images in the test set using the optimized recognition classifier, and completing the training when the recognition classification result exceeds a threshold, otherwise continuing the training.

[0037] According to the system of the second aspect of the present invention, the fourth processing unit is specifically configured to select a signal having the signal duration within the specified bandwidth range of the radio frequency signal and perform the downconversion processing.

[0038] According to the system of the second aspect of the present invention, the sixth processing unit is specifically configured to use the trained recognition classifier to calculate the similarity between the signal time-frequency domain image of the digital intermediate frequency signal and the M×N signal time-frequency domain images, and determine the recognition result of the signal modulation mode of the radio frequency signal to be identified based on the calculation result of the similarity.

[0039] A third aspect of this invention discloses an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the signal modulation pattern recognition method based on time-frequency domain image matching as described in any one of the first aspects of this disclosure.

[0040] A fourth aspect of this invention discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the signal modulation scheme recognition method based on time-frequency domain image matching as described in any one of the first aspects of this disclosure.

[0041] In summary, the technical solution of this invention establishes a database of signals with various known modulation methods, and uses time-frequency analysis to obtain the time-frequency domain images of the signals in the dataset as a signal time-frequency domain image set. Finally, signals within the current frequency band are acquired in real time using a general-purpose signal acquisition device, and the same time-frequency analysis as described above is performed to obtain the time-frequency domain image. Image matching is then performed with the signal time-frequency domain image set, thereby improving the efficiency and accuracy of signal modulation method identification. Attached Figure Description

[0042] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0043] Figure 1 This is a flowchart of a signal modulation scheme identification method based on time-frequency domain image matching according to an embodiment of the present invention;

[0044] Figure 2 This is a time-frequency domain image of a linear frequency modulated signal according to an embodiment of the present invention;

[0045] Figure 3 This is a structural diagram of a specific embodiment of a signal modulation mode recognition system based on time-frequency domain image matching according to an embodiment of the present invention;

[0046] Figure 4 This is a structural diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] The first aspect of this invention discloses a signal modulation scheme identification method based on time-frequency domain image matching, the method comprising (e.g.) Figure 1 As shown):

[0049] Step S1: Obtain a signal dataset. The signal dataset contains M types of signals with different signal systems. Each type contains N signals, for a total of M×N signals.

[0050] Step S2: Perform time-frequency analysis on the M×N signals in the signal dataset to obtain M×N time-frequency power spectrum matrices, and use the M×N video power spectrum matrices to generate M×N signal time-frequency domain images;

[0051] Wherein, the horizontal axis of the signal time-frequency domain image is time, the vertical axis is frequency, and the color or grayscale of the coordinate point represents the signal power at the coordinate point.

[0052] Step S3: Divide the M×N signal time-frequency domain images into a training set and a test set according to the ratio, and use the training set and test set to train and test the neural network-based recognition classifier respectively;

[0053] Step S4: Obtain the received radio frequency signal to be identified, and perform down-conversion processing on the radio frequency signal to obtain a digital intermediate frequency signal;

[0054] Step S5: Perform time-frequency analysis on the digital intermediate frequency signal to obtain the time-frequency power spectral density of the digital intermediate frequency signal, and generate a signal time-frequency domain image of the digital intermediate frequency signal;

[0055] Step S6: Using the trained recognition classifier, the recognition result of the signal modulation mode of the radio frequency signal to be identified is obtained by matching the signal time-frequency domain image of the digital intermediate frequency signal with the M×N signal time-frequency domain images.

[0056] In step S1, a signal dataset is obtained, which contains M types of signals with different signal systems, each type containing N signals, for a total of M×N signals.

[0057] Specifically, signal datasets with different modulation schemes are generated: based on the functional expression of the signal, the sampling rate f is set using simulation software or a standard signal source. s Given the duration τ of each signal segment and the size N of the dataset, we obtain a signal dataset S that traverses different modulation schemes within an adjustable parameter range. i (n), S i (n) is {S i (0) S i (1) … S i (N)}, i=1,2,...,M, where i is the signal modulation method.

[0058] In some embodiments, in step S1, with a specified frequency point as the center frequency point, within a specified bandwidth range, the M×N signals are generated according to the signal function expression, the set sampling rate, the signal duration, and the signal size N of each type of signal to form the signal dataset.

[0059] In some embodiments, the signals with different modulation methods include at least linear frequency modulated signals, single-frequency signals, spread spectrum signals, and pulse signals; the function expressions corresponding to the linear frequency modulated signals, single-frequency signals, pulse signals, and spread spectrum signals in step S1 are respectively...

[0060] Linear frequency modulation signal:

[0061] Single-frequency signal:

[0062] Spread spectrum signal:

[0063] Pulse signal:

[0064] Where A(t) is the signal amplitude, j is the imaginary unit, f0 is the initial carrier frequency, T is the dwell time at a certain frequency point, K is a constant coefficient of frequency change with time, and t is time; A S ,f, Let be the signal amplitude, frequency, and initial phase; C(t) be the spreading code of the carrier modulation; and τ and T be the pulse width and repetition period, respectively.

[0065] In some embodiments, in step S1, when generating N signals for each type of signal modulation method, the N signals for that type of signal modulation method are generated by adjusting the signal parameters. The adjusted signal parameters specifically include:

[0066] For the linear frequency modulated signal, the adjusted signal parameters include the starting frequency, ending frequency, starting power, ending power, and frequency modulation rate;

[0067] For the single-frequency signal, the adjusted signal parameters include carrier frequency and carrier power;

[0068] For the spread spectrum signal, the adjusted signal parameters include the spread spectrum code rate, carrier frequency, and carrier power;

[0069] For the pulse signal, the adjusted signal parameters include pulse period, duty cycle, and carrier power.

[0070] In step S2, time-frequency analysis is performed on each of the M×N signals in the signal dataset to obtain M×N time-frequency power spectrum matrices, and M×N signal time-frequency domain images are generated using the M×N video power spectrum matrices. In the signal time-frequency domain images, the horizontal axis represents time, the vertical axis represents frequency, and the signal power at each coordinate point is represented by its color or grayscale value.

[0071] Specifically, signal time-frequency analysis: for each segment of signal S in the signal dataset... i (n) Perform time-frequency analysis to obtain N×M time-frequency power spectrum matrices S(t,f) and plot them as N×M signal time-frequency domain images. The horizontal axis of the image is time, the vertical axis is frequency, and the color intensity of each point in the image represents the signal power at that point.

[0072] In some embodiments, in step S2, short-time Fourier transforms are performed on the M×N signals in the signal dataset to calculate the M×N time-frequency power spectrum matrices.

[0073] The time-frequency analysis method in step S2 is as follows:

[0074] Dividing a signal into many segments in the time domain and then performing a Fourier transform on each segment is called the Short-Time Fourier Transform (STFT), and its expression is as follows:

[0075] STFT(t,f)=∫s(t′)f(tt′)exp(-jft′)dt′

[0076] The only difference between STFT and FFT (Fast Fourier Transform) lies in the window function f(t). Because the time domain is no longer infinite, it's possible to monitor how the signal spectrum changes as a function of time. STFT can also be represented in the frequency domain, as...

[0077]

[0078] Due to the uncertainty principle, STFTs cannot achieve high measurement accuracy in both time and frequency domains simultaneously; high resolution in the frequency domain implies low resolution in the time domain, and vice versa. Furthermore, the window function also affects STFT performance; commonly used window functions include the Hamming window, Hanning window, Kaiser window, and Gaussian window.

[0079] In step S3, the M×N time-frequency domain images of the signal are divided into a training set and a test set according to the ratio, and the neural network-based recognition classifier is trained and tested using the training set and the test set respectively.

[0080] Specifically, a signal time-frequency domain image set is established: N×M signal time-frequency domain images are stored in M ​​classes, with N images in each class; the signals in each class are divided into two parts according to the proportion to form the training set and test set of the neural network, and finally, M signal classification labels are established.

[0081] In some embodiments, in step S3, training the recognition classifier specifically includes: optimizing the recognition classifier using the training set, using the optimized recognition classifier to recognize images in the test set, completing the training when the recognition and classification result exceeds a threshold, otherwise continuing the training.

[0082] In step S4, the received radio frequency signal to be identified is acquired, and the radio frequency signal is down-converted to obtain a digital intermediate frequency signal.

[0083] Specifically, signal acquisition and down-conversion: An antenna receives radio frequency signals and transmits them to a signal acquisition device. The signal acquisition device down-converts the signal with a duration of τ within a BMHz bandwidth (a specified bandwidth) corresponding to the monitored signal frequency point into a digital intermediate frequency signal S(n), which is then transmitted to a signal recognition device. The antenna, signal acquisition device, and signal processing device are connected sequentially. The signal processing device applies the aforementioned signal modulation scheme recognition method based on time-frequency domain image matching.

[0084] In some embodiments, in step S4, a signal with the signal duration within the specified bandwidth range of the radio frequency signal is selected, and the downconversion process is performed.

[0085] Step S5: Perform time-frequency analysis on the digital intermediate frequency signal to obtain the time-frequency power spectral density of the digital intermediate frequency signal, and generate a signal time-frequency domain image of the digital intermediate frequency signal;

[0086] Specifically, signal time-frequency analysis: The signal identification device reads the digital signal S(n) transmitted by the signal acquisition device, and performs time-frequency analysis using the signal time-frequency analysis method described in step S2 to obtain a signal time-frequency domain image.

[0087] Step S6: Using the trained recognition classifier, the recognition result of the signal modulation mode of the radio frequency signal to be identified is obtained by matching the signal time-frequency domain image of the digital intermediate frequency signal with the M×N signal time-frequency domain images.

[0088] Specifically, signal modulation mode identification: load the signal time-frequency domain image set saved in step S3, perform image matching between the signal time-frequency domain image obtained in step S5 and the signal time-frequency domain image set to obtain the modulation mode classification result.

[0089] In some embodiments, in step S6, the trained recognition classifier is used to calculate the similarity between the signal time-frequency domain image of the digital intermediate frequency signal and the M×N signal time-frequency domain images, and the recognition result of the signal modulation mode of the radio frequency signal to be identified is determined based on the similarity calculation result.

[0090] The image matching algorithm in step S6 is the Normalized Product Correlation (NCC) algorithm. Of course, other image matching algorithms can also be used, such as the Mean Absolute Difference (SAD) algorithm, the Sum of Absolute Errors (SAD) algorithm, etc., and there is no limitation here.

[0091] Specific examples

[0092] Taking the identification of four common signal modulation methods within a 40MHz band centered at 1575MHz as an example, these are linear frequency modulation (LFM) signals, single-frequency signals, pulse signals, and spread spectrum signals. Their functional expressions are shown in Table 1:

[0093] Table 1 Function expressions for several common signal types

[0094]

[0095] Where A(t) is the signal amplitude, j is the imaginary unit, f0 is the initial carrier frequency, T is the dwell time at a certain frequency point, K is a constant coefficient of frequency change with time, and t is time; A S ,f, Let be the signal amplitude, frequency, and initial phase; C(t) be the spreading code of the carrier modulation; and τ and T be the pulse width and repetition period, respectively.

[0096] Of course, Table 1 is only a relatively general expression for four types of signals. Signals can also use other expressions, and signal modulation methods are not limited to linear frequency modulation (LFM), single-frequency (SFM), pulse, and spread spectrum signals. Based on the signal function expressions for the four modulation methods, the sampling rate f is set using simulation software or a standard signal source. sGiven a signal dataset S with a clock frequency of 125MHz (USRP N310 master clock frequency), a signal duration of 81.92μs per segment (corresponding to 10240 sampling points), and a dataset size of N=1500, we obtain the following dataset: i (n), S i (n) is {S i (0) S i (1) … S i (N)}, i=1,2,...,4, where i is the signal modulation method. The dataset should also include the case of normal background noise.

[0097] The standard signal source uses an E4438C vector signal generator. The signal source panel options are set to iterate through signals of different types and parameter ranges. Similarly, signals of different types and parameter ranges are generated through software programming. The adjustable parameter types for each signal are shown in Table 2.

[0098] Table 2 Adjustable parameters for common signal types

[0099]

[0100] Note: Frequency modulation rate = frequency modulation bandwidth / (number of steps × dwell time); Duty cycle = pulse width / pulse period

[0101] The above are only some of the variable signal parameters. The contents of Table 2 do not represent a limitation on the types of variable parameters of the signal.

[0102] For each segment of signal S in the signal dataset i (n) To perform time-frequency domain analysis, firstly, the time-frequency power spectrum matrix S(t,f) of the input signal is calculated using the short-time Fourier transform. The formula for calculating S(t,f) is:

[0103] S(t,f)=∫s(t′)f(tt′)exp(-jft′)dt′

[0104] This signal is then plotted as a time-frequency domain image, with the horizontal axis representing time and the vertical axis representing frequency. The color intensity of each point in the image indicates the signal power at that point. For example... Figure 2 As shown, this displays a time-frequency domain image of a linear frequency modulated signal.

[0105] After time-frequency analysis, two one-dimensional time-frequency domain sequences are obtained. This process is equivalent to preprocessing the signal, extracting signal feature parameters, and enhancing the signal's characteristics.

[0106] The preprocessed time-frequency sequences are stored according to five signal types, with 1500×2 segments in each category. Each category's 1500×2 segments are then divided into two parts at a 9:1 ratio to form the training and testing sets for the neural network. Finally, image classification labels are established for these five signal types: Sweep (linear frequency modulated signal), Pulse (pulse signal), SF (single-frequency signal), DSSS (spread spectrum signal), and Awgn (normal background noise).

[0107] The general-purpose software-defined radio equipment uses the USRP N310, which connects the antenna and the signal identification device—a personal computer—to form a hardware system. The USRP N310 receives radio frequency signals through the antenna, downconverts the monitored signal frequency point corresponding to the 81.92μs signal within a 40MHz bilateral bandwidth into a digital intermediate frequency signal S(n), and transmits it to the computer via a network protocol.

[0108] The computer performs time-frequency analysis on the signal to obtain a time-frequency domain image of the signal, then loads the already saved time-frequency domain image set of the signal, and finally performs image matching between the time-frequency domain image of the signal and the time-frequency domain image set of the signal to obtain the modulation mode classification result.

[0109] A second aspect of this invention discloses a signal modulation scheme recognition system based on time-frequency domain image matching. The system includes:

[0110] The first processing unit is configured to acquire a signal dataset, the signal dataset containing M types of signals with different signal systems, each type containing N signals, for a total of M×N signals;

[0111] The second processing unit is configured to perform time-frequency analysis on the M×N signals in the signal dataset to obtain M×N time-frequency power spectrum matrices, and use the M×N video power spectrum matrices to generate M×N signal time-frequency domain images.

[0112] Wherein, the horizontal axis of the signal time-frequency domain image is time, the vertical axis is frequency, and the color or grayscale of the coordinate point represents the signal power at the coordinate point.

[0113] The third processing unit is configured to divide the M×N signal time-frequency domain images into a training set and a test set according to the ratio, and to train and test a neural network-based recognition classifier using the training set and the test set respectively.

[0114] The fourth processing unit is configured to acquire the received radio frequency signal to be identified and perform down-conversion processing on the radio frequency signal to obtain a digital intermediate frequency signal.

[0115] The fifth processing unit is configured to perform time-frequency analysis on the digital intermediate frequency signal to obtain the time-frequency power spectral density of the digital intermediate frequency signal and generate a signal time-frequency domain image of the digital intermediate frequency signal.

[0116] The sixth processing unit is configured to use the trained recognition classifier to obtain the recognition result of the signal modulation mode of the radio frequency signal to be identified by matching the signal time-frequency domain image of the digital intermediate frequency signal with the M×N signal time-frequency domain images.

[0117] According to the system of the second aspect of the present invention, the first processing unit is specifically configured to, in step S1, generate the M×N signals within a specified bandwidth range, using a specified frequency point as the center frequency point, based on the function expression of the signal, the set sampling rate, the signal duration, and the signal size N of each type of signal, to form the signal dataset.

[0118] According to the system of the second aspect of the present invention, the signals of different signal modulation methods include at least linear frequency modulation signals, single-frequency signals, spread spectrum signals, and pulse signals; the first processing unit is specifically configured to generate N signals of each type of signal modulation method by adjusting signal parameters when generating N signals for each type of signal modulation method, wherein the adjusted signal parameters specifically include:

[0119] For the linear frequency modulated signal, the adjusted signal parameters include the starting frequency, ending frequency, starting power, ending power, and frequency modulation rate;

[0120] For the single-frequency signal, the adjusted signal parameters include carrier frequency and carrier power;

[0121] For the spread spectrum signal, the adjusted signal parameters include the spread spectrum code rate, carrier frequency, and carrier power;

[0122] For the pulse signal, the adjusted signal parameters include pulse period, duty cycle, and carrier power.

[0123] According to the system of the second aspect of the present invention, the second processing unit is specifically configured to perform short-time Fourier transform on each of the M×N signals in the signal dataset to calculate the M×N time-frequency power spectrum matrices.

[0124] According to the system of the second aspect of the present invention, the third processing unit is specifically configured to train the recognition classifier by: optimizing the recognition classifier using the training set, recognizing images in the test set using the optimized recognition classifier, and completing the training when the recognition classification result exceeds a threshold, otherwise continuing the training.

[0125] According to the system of the second aspect of the present invention, the fourth processing unit is specifically configured to select a signal having the signal duration within the specified bandwidth range of the radio frequency signal and perform the downconversion processing.

[0126] According to the system of the second aspect of the present invention, the sixth processing unit is specifically configured to use the trained recognition classifier to calculate the similarity between the signal time-frequency domain image of the digital intermediate frequency signal and the M×N signal time-frequency domain images, and determine the recognition result of the signal modulation mode of the radio frequency signal to be identified based on the calculation result of the similarity.

[0127] A specific implementation of a signal modulation scheme recognition system based on time-frequency domain image matching includes (e.g.) Figure 3 As shown): an antenna (110), a signal acquisition device (120), and a signal processing device (130) are connected in sequence. The signal acquisition device is a software radio device, and the model used is USRP N310. The USRP N310 connects the antenna and the signal identification device—a personal computer—to form a hardware system. The hardware system uses the above method to identify the signal modulation mode.

[0128] A third aspect of this invention discloses an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the signal modulation pattern recognition method based on time-frequency domain image matching as described in any one of the first aspects of this disclosure.

[0129] Figure 4 This is a structural diagram of an electronic device according to an embodiment of the present invention, such as... Figure 4 As shown, the electronic device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, Near Field Communication (NFC), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0130] Those skilled in the art will understand that Figure 4The structure shown is merely a structural diagram of the part related to the technical solution of this disclosure and does not constitute a limitation on the electronic device to which the solution of this application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0131] A fourth aspect of this invention discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the signal modulation scheme recognition method based on time-frequency domain image matching as described in any one of the first aspects of this disclosure.

[0132] In summary, the technical solution of this invention establishes a database of signals with various known modulation schemes. Using time-frequency analysis, the time-frequency domain images of the signals in the dataset are used as a signal time-frequency domain image set. Finally, signals within the current frequency band are acquired in real time using a general-purpose signal acquisition device. After performing the same time-frequency analysis as described above, the resulting time-frequency domain image is matched with the signal time-frequency domain image set, improving the efficiency and accuracy of signal modulation scheme identification. With further improvements in the computing power of USRP and GPU, the speed of time-frequency domain image matching and signal processing can be further increased, enabling the identification of signal modulation schemes with larger bandwidths and a wider range.

[0133] Please note that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. The above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A signal modulation scheme recognition method based on time-frequency domain image matching, characterized in that, The method includes: Step S1: Obtain a signal dataset. The signal dataset contains M types of signals with different signal systems. Each type contains N signals, for a total of M×N signals. Step S2: Perform time-frequency analysis on the M×N signals in the signal dataset to obtain M×N time-frequency power spectrum matrices, and use the M×N time-frequency power spectrum matrices to generate M×N time-frequency domain images of the signals; Step S3: Divide the M×N signal time-frequency domain images into a training set and a test set according to the ratio, and use the training set and test set to train and test the neural network-based recognition classifier respectively; Step S4: Obtain the received radio frequency signal to be identified, and perform down-conversion processing on the radio frequency signal to obtain a digital intermediate frequency signal; Step S5: Perform time-frequency analysis on the digital intermediate frequency signal to obtain the time-frequency power spectral density of the digital intermediate frequency signal, and generate a signal time-frequency domain image of the digital intermediate frequency signal; Step S6: Using the trained recognition classifier, the recognition result of the signal modulation mode of the radio frequency signal to be identified is obtained by matching the signal time-frequency domain image of the digital intermediate frequency signal with the M×N signal time-frequency domain images.

2. The signal modulation scheme recognition method based on time-frequency domain image matching according to claim 1, characterized in that, In step S1, with a specified frequency point as the center frequency point, within a specified bandwidth range, the M×N signals are generated according to the signal function expression, the set sampling rate, the signal duration, and the signal size N of each type of signal to form the signal dataset.

3. The signal modulation scheme recognition method based on time-frequency domain image matching according to claim 2, characterized in that, The signals with different modulation methods include at least linear frequency modulation signals, single-frequency signals, spread spectrum signals, and pulse signals; in step S1, when generating N signals for each type of modulation method, the N signals for that type of modulation method are generated by adjusting the signal parameters. The adjusted signal parameters specifically include: For the linear frequency modulated signal, the adjusted signal parameters include: starting frequency, ending frequency, starting power, ending power, and frequency modulation rate; For the single-frequency signal, the adjusted signal parameters include: carrier frequency and carrier power; For the spread spectrum signal, the adjusted signal parameters include: spreading code rate, carrier frequency, and carrier power; For the pulse signal, the adjusted signal parameters include: pulse period, duty cycle, and carrier power.

4. The signal modulation scheme recognition method based on time-frequency domain image matching according to claim 3, characterized in that, in, The horizontal axis of the signal time-frequency domain image is time, and the vertical axis is frequency. The color or grayscale of the coordinate point represents the signal power at the coordinate point. In step S2, short-time Fourier transform is performed on the M×N signals in the signal dataset to calculate the M×N time-frequency power spectrum matrices.

5. The signal modulation scheme recognition method based on time-frequency domain image matching according to claim 4, characterized in that, In step S3, training the recognition classifier specifically includes: optimizing the recognition classifier using the training set, using the optimized recognition classifier to recognize images in the test set, and completing the training when the recognition and classification result exceeds a threshold; otherwise, continuing the training.

6. The signal modulation scheme recognition method based on time-frequency domain image matching according to claim 5, characterized in that, In step S4, a signal with the specified duration within the specified bandwidth range of the radio frequency signal is selected, and the downconversion process is performed.

7. The signal modulation scheme recognition method based on time-frequency domain image matching according to claim 6, characterized in that, In step S6, the trained recognition classifier is used to calculate the similarity between the signal time-frequency domain image of the digital intermediate frequency signal and the M×N signal time-frequency domain images, and the recognition result of the signal modulation mode of the radio frequency signal to be identified is determined based on the similarity calculation result.

8. A signal modulation scheme recognition system based on time-frequency domain image matching, characterized in that, The system includes: The first processing unit is configured to acquire a signal dataset, the signal dataset containing M types of signals with different signal systems, each type containing N signals, for a total of M×N signals; The second processing unit is configured to perform time-frequency analysis on the M×N signals in the signal dataset to obtain M×N time-frequency power spectrum matrices, and use the M×N time-frequency power spectrum matrices to generate M×N time-frequency domain images of the signals. The third processing unit is configured to divide the M×N signal time-frequency domain images into a training set and a test set according to the ratio, and to train and test a neural network-based recognition classifier using the training set and the test set respectively. The fourth processing unit is configured to acquire the received radio frequency signal to be identified and perform down-conversion processing on the radio frequency signal to obtain a digital intermediate frequency signal. The fifth processing unit is configured to perform time-frequency analysis on the digital intermediate frequency signal to obtain the time-frequency power spectral density of the digital intermediate frequency signal and generate a signal time-frequency domain image of the digital intermediate frequency signal. The sixth processing unit is configured to use the trained recognition classifier to obtain the recognition result of the signal modulation mode of the radio frequency signal to be identified by matching the signal time-frequency domain image of the digital intermediate frequency signal with the M×N signal time-frequency domain images.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the steps of the signal modulation mode recognition method based on time-frequency domain image matching according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the signal modulation mode recognition method based on time-frequency domain image matching according to any one of claims 1 to 7.

Images