Modulation identification method and device based on microwave photon frequency-time mapping

Abstract

The invention discloses a modulation identification method based on microwave photon frequency-time mapping. The method comprises the following steps: firstly, performing microwave photon frequency-time mapping on a to-be-measured microwave signal, and mapping frequency information of the to-be-measured microwave signal to a time domain to obtain a time-frequency characteristic of the to-be-measured microwave signal; and inputting the time-frequency characteristics of the to-be-detected microwave signal into a pre-trained neural network model so as to identify the modulation form of the to-be-detected microwave signal. The invention also discloses a modulation identification device based on microwave photon frequency-time mapping. Compared with the prior art, on the basis of a microwave photon frequency-time mapping model, Fourier transform is directly completed through an optical device, a radio frequency signal is modulated to be processed in an optical domain, time delay characteristics of dispersion broadening and compression are combined, the picosecond-level time stretching precision is achieved, the instantaneous bandwidth limitation of a traditional spectrum analyzer is broken through, and the spectrum analysis precision is improved. Frequency spectrum processing of large-bandwidth signals can be completed in real time, computing resources are saved, and the real-time performance of modulation identification is improved.

Description

Technical Field

[0001] The present invention relates to a modulation recognition method for recognizing the modulation form of microwave signals. Background Art

[0002] Wireless communication technology has profoundly changed our lives. The traditional spectrum allocation method is based on frequency, independently allocated with clear boundaries; while the new generation of communication adopts the method of dynamic spectrum sharing, which improves the spectrum utilization efficiency and also increases the difficulty of spectrum supervision. Facing the new requirements and challenges of future communication, we need to accurately identify relevant parameters including the signal modulation form to ensure the security and stability of communication. So-called modulation recognition refers to judging the modulation form adopted by a signal on the premise of unknown modulation information content and modulation parameters. Existing modulation recognition technologies usually obtain the time-frequency characteristics of microwave signals through time-frequency mapping and analysis of microwave signals, and then realize the recognition of their modulation forms according to the time-frequency characteristics of the signals. The time-frequency diagram of a signal is an image that visualizes the changes of the signal in time and frequency. The horizontal axis of the time-frequency diagram represents time, the vertical axis represents frequency, and the brightness or color of each point represents the intensity or energy of the signal at that moment and frequency. With the increasing development of machine learning, some technical solutions that combine time-frequency diagrams with artificial neural networks to achieve modulation recognition have emerged currently.

[0003] Existing technologies usually realize the time-frequency mapping and analysis of microwave signals in the electrical domain. mainly through digital signal processing technology, converting the analog signal to be measured into a digital signal through analog-to-digital conversion, and obtaining the time-frequency characteristics of the microwave signal through processing methods such as fast Fourier transform, wavelet transform or Hilbert transform in the digital domain. These methods decompose the signal into a series of time-frequency atoms in time-frequency and represent them in the form of a time-frequency diagram. The core of these solutions is to calculate the spectrum frame by frame and then combine to generate a time-frequency diagram. Its calculation is mainly implemented by a high-performance digital processor or FPGA, and the processing speed is limited by the computing power of electronic hardware and the data transmission bandwidth, and the calculation amount is large when analyzing ultra-wideband signals. Summary of the Invention

[0004] The purpose of the present invention is to provide a modulation recognition method based on microwave photonic frequency-time mapping, and realize the time-frequency mapping and analysis of microwave signals based on microwave photonic frequency-time mapping, so as to overcome the problems existing in the traditional electrical domain time-frequency mapping and analysis methods, such as insufficient instantaneous analysis bandwidth, large noise, poor real-time performance, large calculation amount, etc., and be able to realize real-time and accurate recognition of various types of signals such as linear frequency modulation signals, frequency hopping signals, and stepped frequency hopping signals.

[0005] The present invention specifically adopts the following technical solutions to solve the above technical problems:

[0006] A modulation recognition method based on microwave photon frequency-time mapping. First, perform microwave photon frequency-time mapping on the microwave signal to be measured, map the frequency information of the microwave signal to be measured to the time domain, and obtain the time-frequency characteristics of the microwave signal to be measured. Then, input the time-frequency characteristics of the microwave signal to be measured into a pre-trained neural network model to identify the modulation form of the microwave signal to be measured.

[0007] In one embodiment, the method of microwave photon frequency-time mapping is specifically as follows: perform time-domain broadening on the optical pulse signal, then perform electro-optic modulation on the time-domain broadened optical pulse signal with the microwave signal to be measured, then perform time-domain compression on the generated modulated optical pulse signal, convert the time-domain compressed modulated optical pulse signal into an electrical signal, and finally obtain the time-frequency diagram of the microwave signal to be measured through digital sampling of the electrical signal. The neural network model is an electrical neural network with the time-frequency diagram as the input and the modulation form as the output.

[0008] In another embodiment, the method of microwave photon frequency-time mapping is specifically as follows: perform time-domain broadening on the optical pulse signal; then perform electro-optic modulation on the time-domain broadened optical pulse signal with the microwave signal to be measured; then divide the generated modulated optical pulse signal into two paths, one path filters out the optical carrier and one sideband and only retains one first-order sideband on one side, and the other path filters out both sidebands and only retains the optical carrier; then perform time-domain compression on the two obtained optical signals respectively and then perform balanced photoelectric detection; finally obtain the time-frequency diagram of the microwave signal to be measured through digital sampling of the obtained electrical signal. The neural network model is an electrical neural network with the time-frequency diagram as the input and the modulation form as the output.

[0009] In yet another embodiment, the method of microwave photon frequency-time mapping is specifically as follows: perform time-domain broadening on the optical pulse signal, then perform electro-optic modulation on the time-domain broadened optical pulse signal with the microwave signal to be measured, then perform time-domain compression on the generated modulated optical pulse signal, and the time-domain compressed modulated optical pulse signal is the time-frequency characteristic of the microwave signal to be measured. The neural network model is an optical neural network with the time-domain compressed modulated optical pulse signal as the input and the modulation form as the output.

[0010] Preferably, positive and negative dispersion media with the same dispersion amount are used for time-domain broadening and time-domain compression respectively.

[0011] Based on the same inventive concept, the following technical solutions can also be obtained:

[0012] A modulation recognition device based on microwave photon frequency-time mapping, comprising:

[0013] A microwave photon frequency-time mapping module, configured to perform microwave photon frequency-time mapping on the microwave signal to be measured, map the frequency information of the microwave signal to be measured to the time domain, and obtain the time-frequency characteristics of the microwave signal to be measured;

[0014] A neural network model, pre-trained, is used to take the time-frequency features of the microwave signal to be measured as input and output the modulation form of the microwave signal to be measured.

[0015] In one embodiment, the microwave photonic frequency-time mapping module includes:

[0016] A time-domain broadening module for broadening the optical pulse signal in the time domain;

[0017] An electro-optic modulation module for electro-optically modulating the time-domain broadened optical pulse signal with the microwave signal to be measured;

[0018] A time-domain compression module for compressing the modulated optical pulse signal generated by the electro-optic modulation module in the time domain; a photodetector for converting the time-domain compressed modulated optical pulse signal into an electrical signal;

[0019] A digital processing module for obtaining the time-frequency diagram of the microwave signal to be measured by digitally sampling the electrical signal;

[0020] The neural network model is an electrical neural network with the time-frequency diagram as input and the modulation form as output.

[0021] In another embodiment, the microwave photonic frequency-time mapping module includes:

[0022] A time-domain broadening module for broadening the optical pulse signal in the time domain;

[0023] An electro-optic modulation module for electro-optically modulating the time-domain broadened optical pulse signal with the microwave signal to be measured;

[0024] A power splitter for splitting the generated modulated optical pulse signal into two paths;

[0025] A first optical filter for filtering out the optical carrier and one sideband and only retaining one side first-order sideband;

[0026] A second optical filter for filtering out both sidebands of the other path and only retaining the optical carrier;

[0027] Two time-domain compression modules for respectively compressing the two paths of optical signals after filtering in the time domain;

[0028] A balanced photodetector for performing balanced photodetection on the two paths of optical signals after time-domain compression;

[0029] A digital processing module for obtaining the time-frequency diagram of the microwave signal to be measured by digitally sampling the electrical signal output by the balanced photodetector;

[0030] The neural network model is an electrical neural network with the time-frequency diagram as input and the modulation form as output.

[0031] In yet another embodiment, the microwave photonic frequency-time mapping module includes:

[0032] A time-domain broadening module for broadening the optical pulse signal in the time domain;

[0033] An electro-optic modulation module for electro-optically modulating the time-domain broadened optical pulse signal with the microwave signal to be measured;

[0034] A time-domain compression module for compressing the modulated optical pulse signal generated by the electro-optic modulation module in the time domain, and the time-domain compressed modulated optical pulse signal is the time-frequency characteristic of the microwave signal to be measured;

[0035] The neural network model is an optical neural network that takes the time-domain compressed modulated optical pulse signal as the input and the modulation form as the output.

[0036] Preferably, the time-domain broadening module and the time-domain compression module are positive and negative dispersion media with the same dispersion amount, respectively.

[0037] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0038] Based on the microwave photonic frequency-time mapping model, the present invention directly completes the Fourier transform through optical devices, modulates the radio frequency signal to the optical domain for processing, combines the time-delay characteristics of dispersion broadening and compression, realizes picosecond-level time stretching accuracy, breaks through the instantaneous bandwidth limitation of traditional spectrum analyzers, can complete the spectrum processing of large-bandwidth signals in real time, saves computing resources, and improves the real-time performance of modulation recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a principle block diagram of the technical solution of the present invention;

[0040] Figure 2 is a structural schematic diagram of the first specific embodiment of the present invention;

[0041] Figure 3 is a structural schematic diagram of the second specific embodiment of the present invention;

[0042] Figure 4 is a structural schematic diagram of the third specific embodiment of the present invention;

[0043] Figure 5 is an identification effect diagram of the present invention for a triangular frequency modulation signal. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] Aiming at the deficiencies of existing modulation recognition technologies, the solution idea of the present invention is based on a microwave photon frequency-time mapping model. By directly performing Fourier transform through optical devices, the radio frequency signal is modulated to the optical domain for processing. Combining the time-delay characteristics of dispersion broadening and compression, the instantaneous bandwidth limitation of traditional spectrum analysis is broken through, while computational resources are saved and the real-time performance of modulation recognition is improved.

[0045] The modulation recognition device based on microwave photon frequency-time mapping proposed by the present invention, as Figure 1 shown, includes:

[0046] A microwave photon frequency-time mapping module, which is used to perform microwave photon frequency-time mapping on the microwave signal to be measured, map the frequency information of the microwave signal to be measured to the time domain, and obtain the time-frequency characteristics of the microwave signal to be measured;

[0047] A neural network model, which is pre-trained and used to take the time-frequency characteristics of the microwave signal to be measured as input and output the modulation form of the microwave signal to be measured.

[0048] For the convenience of public understanding, the technical solution of the present invention will be described in detail through several specific embodiments below:

[0049] Embodiment 1

[0050] The structure of the modulation recognition device in this embodiment is as Figure 2 shown. The optical pulse signal emitted by the laser is dispersion-stretched by the dispersion medium 1 (with positive dispersion amount) to achieve the time-domain broadening of the optical pulse; the Mach-Zehnder modulator (MZM) modulates the microwave signal to be measured onto the time-domain broadened optical pulse signal to generate a modulated optical pulse signal containing a carrier and positive and negative first-order sidebands; the modulated optical pulse signal output by the MZM is then dispersion-compressed by the dispersion medium 2 (with negative dispersion amount and the same dispersion amount as that of the dispersion medium 1) to achieve the time-domain compression of the optical pulse, and the frequency information of the microwave signal to be measured is converted to the time domain; the time-domain compressed modulated optical pulse signal is converted into an electrical signal by a photodetector; the digital processing module obtains the time-frequency diagram of the microwave signal to be measured by digitally sampling the electrical signal; inputting the time-frequency diagram into the pre-trained electrical neural network model can output the modulation form of the microwave signal to be measured.

[0051] For the convenience of public understanding, the basic principle of this embodiment will be described below:

[0052] In a photon frequency measurement system, the research on the dispersion medium mainly focuses on group velocity dispersion, mainly focusing on β2. D1 and D2 represent positive and negative dispersion media respectively. In the frequency domain, the functions of D1 and D2 are to introduce phase shifts to different frequencies, and the introduced phase shifts have a quadratic relationship with the frequency. Using It is expressed that the second derivative of the dispersion medium mode propagation constant β0 with respect to frequency is obtained through Fourier transform. The time-domain form of its response is:

[0053]

[0054] The pulsed signal output by the laser is denoted as u0(t). After being dispersion-stretched by D1 and broadened in the time domain, the output of D1 is denoted as u1(t), where denotes the convolution symbol, and there is:

[0055]

[0056]

[0057] Since the pulse width of the laser output is very narrow and the large dispersion of the system can meet the requirements of the far-field time-domain condition, that is, the pulse signal width is much smaller than the dispersion, the τ 2 term can be directly ignored. The original formula can be simplified to:

[0058]

[0059] It can be seen that the role of D1 is to broaden the pulse. After the dispersion stretching module of D1, the RF electrical signal to be measured needs to be loaded on the optical signal through the electro-optic modulation of MZM, and the output optical signal is u2(t). The modulation relationship can be written as:

[0060] u2(t)≈u1(t)×[a + bβ m s(t)]

[0061] Output signal u3(t):

[0062]

[0063] Substituting the modulation relationship, we have:

[0064]

[0065] The au0(t) term is the initial pulse component input by the laser, The term represents that the system introduces a quadratic relationship with the phase changing with time to the output signal, that is, frequency chirping. S(ω) is the spectrum of the RF signal to be measured input, and ω corresponds to

[0066] When the signal to be measured is a single-frequency signal cos(ωt) with a frequency of f,

[0067]

[0068] In the above formula, in the time domain, 0, There is a pulse signal at each moment. After detection by the photodetector, it does not affect the time-domain positions of the three pulses. The frequency-time mapping relationship can be obtained as follows:

[0069]

[0070]

[0071] After photoelectric conversion, information is collected by a data acquisition card. A single output waveform can be expressed as:

[0072] I(t) = ax0(t) + b -1 x0(t + 2πf k φ) + b1x0(t - 2πf k φ)

[0073] It can be seen that after microwave photon frequency-time mapping, the system outputs three pulses in the time domain. The middle one is the original reference optical pulse, and the two on both sides are symmetric pulse signals. There is a certain relationship between the time interval between the pulse signal and the original optical pulse signal and the frequency value. The frequency of the microwave signal to be measured can be deduced according to the time interval of the time-domain pulse.

[0074] Through the above processing, the signal is divided into multiple small time windows, and each window contains a segment of the signal. By arranging the Fourier transform results of each time window on the time axis, a two-dimensional time-frequency distribution diagram (time-frequency diagram) can be generated. The horizontal axis represents time, and the vertical axis represents frequency. The generated time-frequency diagram is sent into a pre-trained electrical neural network model according to different analysis requirements. Based on deep learning-driven time-frequency analysis, the time-frequency diagram features are extracted by the electrical neural network for signal decomposition or classification. By extracting eigenvalues such as convolution on the image, the extracted features are input into the subsequent calculation layer to output the classification result or prediction value.

[0075] Embodiment 2

[0076] The structure of the modulation recognition device in this embodiment is as Figure 3As shown in the figure, the optical pulse signal emitted by the laser is dispersion-stretched by the dispersion medium 1 (whose dispersion amount is positive), so as to realize the time-domain broadening of the optical pulse; the Mach-Zehnder modulator (MZM) modulates the microwave signal to be measured onto the time-domain broadened optical pulse signal, generating a modulated optical pulse signal containing a carrier and positive and negative first-order sidebands; the modulated optical pulse signal is divided into two paths by a power splitter, one path passes through an optical filter 1 to filter out the optical carrier and one sideband and only retains one first-order sideband, and the other path passes through an optical filter 2 to filter out both sidebands and only retains the optical carrier; the two filtered optical signals are respectively time-domain compressed by two identical dispersion media 2, and then subjected to balanced photoelectric detection by a balanced photodetector; the digital processing module obtains the time-frequency diagram of the microwave signal to be measured by digitally sampling the electrical signal output by the balanced photodetector; by inputting the time-frequency diagram into a pre-trained electrical neural network model, the modulation form of the microwave signal to be measured can be output.

[0077] Taking the microwave frequency-hopping signal to be measured as an example, the expression of the signal to be measured is as follows:

[0078]

[0079] where \(0\leq t\leq T\), \(T\) is the total observation time, \(K\) is the total number of hops within the observation time, \(T_H\) is the hop period, \(a(t)\) is the complex envelope of the baseband, \(A\) s is the amplitude of the frequency-hopping signal, \(f\) k and are respectively the carrier frequency and phase of the \(k\)-th hop.

[0080] In the photon frequency measurement system, the research on the dispersion medium mainly focuses on group velocity dispersion. Mainly concerned with \(\beta_2\), \(D_1\) and \(D_2\) represent positive and negative dispersion media respectively. The propagation functions of \(D_1\) and \(D_2\) in the frequency domain and their corresponding time-domain forms are:

[0081]

[0082] Denote the pulse signal output by the laser as \(u_0(t)\). After being dispersion-stretched by \(D_1\) and time-domain broadened, the output of \(D_1\) is denoted as \(u_1(t)\), where represents the convolution symbol, and there is:

[0083]

[0084]

[0085] Since the pulse width of the laser output is very narrow, and the large dispersion amount of the system can meet the requirements of the far-field time-domain condition, that is, the pulse signal width is much smaller than the dispersion amount, the \(\tau\) 2 term can be directly ignored. The original formula can be simplified as:

[0086]

[0087] It can be seen that the function of D1 is to perform pulse broadening. After passing through the dispersion stretching module of D1, the radio frequency electrical signal to be measured needs to be loaded onto the optical signal through the electro-optic modulation of MZM, and the output optical signal is u2(t). The modulation relationship can be written as:

[0088] u2(t)≈u1(t)×[a + bβ m s(t)]

[0089] Output signal u3(t):

[0090]

[0091] After filtering out the optical carrier and the -1st order sideband in the upper branch, it is denoted as u 3(1) (t), and after filtering out the ±1st order sidebands in the lower branch, it is denoted as u 3(2) (t).

[0092] Substituting into the modulation relationship, we get:

[0093]

[0094] The term au0(t) is the initial pulse component input by the mode-locked laser,

[0095] The term represents that the system introduces a quadratic relationship with the phase changing over time to the output signal, that is, frequency chirping. S(ω) is the spectrum of the RF signal to be measured input, and ω corresponds to

[0096] When the signal to be measured is a single-frequency signal cos(ωt) with a frequency of f,

[0097]

[0098] u 3(2) (t) = au0(t)

[0099] Simplify u 3(1) (t) and u 3(2) (t):

[0100]

[0101] After passing through the 3dB coupler, it is divided into two optical waves u 3(11) (t), and divided into two optical waves u 3(21) (t),

[0102] Due to the use of the 3dB coupler, the phases of the two detected optical waves differ by π / 2. u 3(11) (t) is detected by the photodetector 1, u 3(21)(t) is detected by the photodetector 2. Since the detector is a square-law detector, the current signals output by the two detectors are i1(t) and i2(t) respectively:

[0103]

[0104] After being amplified by the differential amplifier, part of the noise and the DC term bias are cancelled, reducing the noise of the system.

[0105] At time 0 in the time domain, there is a pulse signal at each:

[0106]

[0107]

[0108] The microwave signal is modulated onto the optical carrier by the modulator. In the optical domain, the frequency information of the microwave signal is mapped to the time domain through frequency-time mapping. After beat frequency detection by the photodetector, the information can be collected by a data acquisition card and then recognized by an electrical neural network. When using a data acquisition card for information collection, the single output waveform on the display can be expressed as:

[0109] I(t) = ax0(t) + b1x0(t - 2πf k φ)

[0110] It can be seen that after microwave photon frequency-time mapping and balanced detection, the system outputs two pulses in the time domain, one is the original reference optical pulse and the other is the pulse signal. There is a certain relationship between the time interval between the pulse signal and the original optical pulse signal and the frequency value. The frequency of the microwave signal to be measured can be deduced according to the time interval of the time-domain pulse, associating the frequency information with the time dimension to achieve frequency-time mapping and generate a two-dimensional time-frequency diagram.

[0111] The time-frequency diagram can be further normalized, denoised and enhanced, and the image resolution can be adjusted according to the input requirements of the electrical neural network, and then input into the electrical neural network for modulation form recognition to output the modulation form of the signal.

[0112] Compared with Embodiment 1, although the microwave photon frequency-time mapping module of this embodiment adds components such as an optical filter and a power splitter, due to the adoption of balanced photodetection, the system noise of the optical link can be effectively eliminated, improving the accuracy of modulation form recognition.

[0113] Embodiment 3

[0114] In the above two embodiments, the optical signal after microwave photonic time-frequency mapping is converted into the electrical domain, and then the time-frequency diagram is extracted through digital processing. Finally, an electrical neural network with the time-frequency diagram as the input and the modulation form as the output is used to identify the modulation form. In fact, the optical signal after microwave photonic time-frequency mapping already contains the complete time-frequency information of the microwave signal to be measured, and the optical neural network can be directly used to extract its features and identify the modulation form. The modulation recognition device of this embodiment is based on this idea, and its structure is as shown in Figure 4 shown. The optical pulse signal emitted by the laser is dispersion-stretched by the dispersion medium 1 (whose dispersion amount is positive), so as to realize the time-domain broadening of the optical pulse; the Mach-Zehnder modulator (MZM) modulates the microwave signal to be measured onto the time-domain broadened optical pulse signal to generate a modulated optical pulse signal containing a carrier and positive and negative first-order sidebands; the modulated optical pulse signal output by the MZM is then dispersion-compressed by the dispersion medium 2 (whose dispersion amount is negative and the same as that of the dispersion medium 1), so as to realize the time-domain compression of the optical pulse, and the frequency information of the microwave signal to be measured is converted into the time domain; the modulated optical pulse signal after time-domain compression is input into the pre-trained optical neural network model, and the modulation form of the microwave signal to be measured can be output.

[0115] The modulation recognition device of Embodiment 2 is used to identify the triangular frequency modulation signal, and the obtained time-frequency diagram is as shown in Figure 5 shown, and it can be seen that the device can accurately identify the triangular frequency modulation signal.

Claims

1. A modulation recognition method based on microwave photon frequency-time mapping, characterized in that: First, microwave photon frequency-time mapping is performed on the microwave signal to be tested, and the frequency information of the microwave signal to be tested is mapped to the time domain to obtain the time-frequency characteristics of the microwave signal to be tested; then the time-frequency characteristics of the microwave signal to be tested are input into a pre-trained neural network model to identify the modulation form of the microwave signal to be tested.

2. The modulation identification method based on microwave photon frequency-time mapping as claimed in claim 1, characterized in that: The microwave photon frequency-time mapping method is specifically as follows: time-domain broadening is performed on an optical pulse signal, and then the time-domain broadened optical pulse signal is electro-optically modulated with a microwave signal to be measured, and then the generated modulated optical pulse signal is time-domain compressed, and the modulated optical pulse signal after time-domain compression is converted into an electrical signal, and finally a time-frequency diagram of the microwave signal to be measured is obtained by digitally sampling the electrical signal; the neural network model is an electrical neural network with the time-frequency diagram as input and a modulated form as output.

3. The modulation identification method based on microwave photon frequency-time mapping as claimed in claim 1, characterized in that: The microwave photon frequency-time mapping method is specifically as follows: time-domain broadening is performed on the optical pulse signal; then the optical pulse signal after time-domain broadening is electro-optically modulated by the microwave signal to be measured; then the generated modulated optical pulse signal is divided into two paths, one path is filtered out of the optical carrier and one side sideband and only one side first-order sideband is retained, and the other path is filtered out of the two sidebands and only the optical carrier is retained; then the obtained two optical signals are respectively compressed in the time domain and then balanced photoelectric detection is performed; finally, the time-frequency diagram of the microwave signal to be measured is obtained by digitally sampling the obtained electrical signal; the neural network model is an electrical neural network with the time-frequency diagram as input and the modulation form as output.

4. The modulation identification method based on microwave photon frequency-time mapping as claimed in claim 1, characterized in that: The microwave photon frequency-time mapping method is specifically as follows: the optical pulse signal is time-domain stretched, and then the time-domain stretched optical pulse signal is electro-optically modulated with the microwave signal to be measured, and then the generated modulated optical pulse signal is time-domain compressed, and the modulated optical pulse signal after time-domain compression is the time-frequency feature of the microwave signal to be measured; the neural network model is an optical neural network with the modulated optical pulse signal after time-domain compression as input and a modulated form as output.

5. The modulation identification method based on microwave photon frequency-time mapping according to any one of claims 2 to 4, characterized in that: Positive and negative dispersion media with the same dispersion amount are used to perform time domain broadening and time domain compression respectively.

6. A modulation identification device based on microwave photon frequency-time mapping, characterized in that: include: The microwave photon frequency-time mapping module is used to perform microwave photon frequency-time mapping on the microwave signal to be measured, map the frequency information of the microwave signal to be measured into the time domain, and obtain the time-frequency characteristics of the microwave signal to be measured; The neural network model is pre-trained and is used to take the time-frequency characteristics of the microwave signal to be measured as input and output the modulation form of the microwave signal to be measured.

7. The modulation identification device based on microwave photon frequency-time mapping as claimed in claim 6, characterized in that: The microwave photon frequency-time mapping module comprises: A time domain stretching module, used for performing time domain stretching on an optical pulse signal; An electro-optical modulation module, used for electro-optically modulating the optical pulse signal after time domain broadening using the microwave signal to be measured; A time domain compression module, used for performing time domain compression on the modulated optical pulse signal generated by the electro-optical modulation module; A photodetector, used for converting the modulated optical pulse signal after time domain compression into an electrical signal; A digital processing module, used for obtaining a time-frequency diagram of the microwave signal to be measured by digitally sampling the electrical signal; The neural network model is an electrical neural network that takes a time-frequency diagram as input and outputs in a modulated form.

8. The modulation identification device based on microwave photon frequency-time mapping as claimed in claim 6, characterized in that: The microwave photon frequency-time mapping module comprises: A time domain stretching module, used for performing time domain stretching on an optical pulse signal; An electro-optical modulation module, used for electro-optically modulating the optical pulse signal after time domain broadening using the microwave signal to be measured; A power divider, used for dividing the generated modulated optical pulse signal into two paths; The first optical filter is used to filter out the optical carrier and one side sideband and retain only one side first-order sideband; The second optical filter is used for filtering out the sidebands on both sides of the other path and retaining only the optical carrier; Two time domain compression modules, used for respectively performing time domain compression on the two optical signals after filtering; A balanced photoelectric detector, used for balanced photoelectric detection of two optical signals after time domain compression; A digital processing module, used for obtaining a time-frequency diagram of the microwave signal to be measured by digitally sampling the electrical signal output by the balanced photoelectric detector; The neural network model is an electrical neural network that takes a time-frequency diagram as input and outputs in a modulated form.

9. The modulation identification device based on microwave photon frequency-time mapping as claimed in claim 6, characterized in that: The microwave photon frequency-time mapping module comprises: A time domain stretching module, used for performing time domain stretching on an optical pulse signal; An electro-optical modulation module, used for electro-optically modulating the optical pulse signal after time domain broadening using the microwave signal to be measured; The time domain compression module is used to perform time domain compression on the modulated optical pulse signal generated by the electro-optical modulation module. The modulated optical pulse signal after time domain compression is the time-frequency characteristic of the microwave signal to be measured. The neural network model is an optical neural network that takes the modulated optical pulse signal after time domain compression as input and outputs in a modulated form.

10. The modulation identification device based on microwave photon frequency-time mapping according to any one of claims 7 to 9, characterized in that: The time domain broadening module and the time domain compression module are respectively positive and negative dispersion media with the same dispersion amount.

Images