Radar signal identification method and device based on photon parameter mapping

Abstract

The invention discloses a radar signal identification method based on photon parameter mapping. The method comprises the following steps: S1, generating two paths of homologous optical carriers; s2, processing the first path of optical carrier into a linear frequency modulation optical signal; performing intensity modulation on the second path of optical carrier wave by using the radar signal to be measured, and introducing delay inequality between two cross-polarization components of the generated modulation optical signal; s3, after the two paths of optical signals are combined, balanced photoelectric detection is carried out, and electrical filtering and envelope detection are carried out in sequence; s4, inputting the data into a modulation type identification module in a digital domain for modulation type identification; inputting the obtained modulation type time sequence into a state recognition module for state recognition; the modulation type identification module and the state identification module are pre-trained deep learning models. The invention also discloses a radar signal identification device based on photon parameter mapping. According to the method, the radar signal modulation type and state can be accurately and rapidly identified through low-speed analog-to-digital conversion.

Description

Technical Field

[0001] The present invention relates to a radar signal recognition method, and belongs to the technical field where radar technology and microwave photon technology intersect. Background Art

[0002] In the modern electromagnetic battlefield, intelligent recognition of radar modulation modes and states is crucial for electronic reconnaissance and jamming. To address the increasing complexity of battlefield electromagnetic environments, the proliferation of stealth aircraft, and the intensifying threat of ultra-low-altitude weapons, traditional radars have gradually evolved into multifunction radars, becoming intelligent perception platforms capable of multi-state operation, mission adaptability, and dynamic reconfiguration. In search and tracking missions, multifunction radars often employ randomly combined pulse parameters (such as pulse width, pulse repetition interval, and carrier frequency), resulting in complex and variable signal forms. This poses significant challenges to traditional recognition techniques based on template matching or statistical histograms. Furthermore, traditional recognition methods typically utilize electronic techniques to process and identify radar signals in the digital domain. However, in broadband recognition scenarios covering a frequency range of 1-40 GHz, the sheer volume of data presents challenges, making ultra-wideband and real-time processing difficult to achieve. This in turn limits system processing efficiency and response speed.

[0003] Microwave photonics, with its ultra-wide bandwidth and inherent parallelism, offers a new solution for rapid feature extraction from broadband signals. Combining microwave photonics with artificial intelligence (AI) promises to enable ultra-wideband and real-time parameter measurement and state recognition. Previous research (F.N. Khan, K. Zhong, X. Zhou, W.H. Al-Arashi, C. Yu, C. Lu, and A.P.T. Lau, “Joint OSNR monitoring and modulation format identification in digital coherent receivers using deep neural networks,” Opt. Express, vol. 25, no. 15, pp. 17767–17776, 2017) has proposed using optoelectronic coherent detection for modulation recognition. Specifically, the received radar signal is modulated onto an optical carrier, then subjected to an optical interferometer structure to generate time-delayed signal replicas. These replicas are then converted to electrical signals via photodetectors. Frequency and phase modulation information is mapped to amplitude variations in the output envelope, which are then combined with deep learning algorithms for modulation format recognition. However, existing solutions are limited to modulation type recognition and still rely on high-speed analog-to-digital converters for data acquisition. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the existing technology and provide a radar signal recognition method based on photon parameter mapping. Through the photon parameter mapping architecture, the frequency, phase and time information of the high-speed radar signal are mapped to the intensity and time information of the low-speed output pulse signal, so that the modulation type and state of the radar signal can be accurately and quickly identified using a low-speed analog-to-digital converter.

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

[0006] A radar signal recognition method based on photon parameter mapping comprises the following steps:

[0007] S1, generate two optical carriers of the same source;

[0008] S2. Processing the first optical carrier into a linear frequency modulated optical signal; intensity modulating the second optical carrier with the radar signal to be measured, and introducing a time delay difference between the two orthogonal polarization components of the generated modulated optical signal to obtain a time domain differential interference optical signal;

[0009] S3. Combining the linear frequency modulated optical signal with the time-domain differential interference optical signal, performing balanced photoelectric detection, and electrically filtering the resulting electrical signal with an electrical filter and then performing envelope detection, thereby obtaining an electrical signal having time-domain pulse characteristics; the electrical filter is a narrowband filter having a passband center frequency within the frequency difference between the linear frequency modulated optical signal and the radar signal to be measured;

[0010] S4. Convert the electrical signal with time-domain pulse characteristics into a digital signal, and input the digital signal into a modulation type recognition module to identify the modulation type of the radar signal to be measured; then input the obtained modulation type time series into a state recognition module to identify the state of the radar signal to be measured; the modulation type recognition module and the state recognition module are both pre-trained deep learning models.

[0011] Further preferably, the following method is used to process the first optical carrier into a linear frequency modulated optical signal: intensity modulating the first optical carrier with an electrical signal whose amplitude varies linearly and periodically; then injecting the generated modulated optical signal into a semiconductor slave laser to generate an optical chirp signal; and finally optically filtering the optical chirp signal to retain only a single-sided first-order sideband, thereby obtaining the linear frequency modulated optical signal.

[0012] Based on the same inventive concept, the following technical solutions can also be obtained:

[0013] A radar signal recognition device based on photon parameter mapping, comprising:

[0014] An optical carrier generation module, used to generate two optical carriers of the same source;

[0015] A linear frequency modulated optical signal generation module, configured to process the first optical carrier into a linear frequency modulated optical signal;

[0016] A time-domain differential interference optical signal generation module is used to modulate the intensity of the second optical carrier with the radar signal to be measured, and introduce a time delay difference between the two orthogonal polarization components of the generated modulated optical signal to obtain a time-domain differential interference optical signal;

[0017] A photoelectric detection and filtering module is used to combine the linear frequency modulated optical signal with the time domain differential interference optical signal, perform balanced photoelectric detection, and use an electrical filter to electrically filter the resulting electrical signal and then perform envelope detection to obtain an electrical signal with time domain pulse characteristics; the electrical filter is a narrowband filter with a passband center frequency within the frequency difference range between the linear frequency modulated optical signal and the radar signal to be measured;

[0018] The digital processing module is used to convert the electrical signal with time-domain pulse characteristics into a digital signal, and input the digital signal into a modulation type recognition module to identify the modulation type of the radar signal to be measured; and then input the obtained modulation type time series into a state recognition module to identify the state of the radar signal to be measured; the modulation type recognition module and the state recognition module are both pre-trained deep learning models.

[0019] Further preferably, the following method is used to process the first optical carrier into a linear frequency modulated optical signal: intensity modulating the first optical carrier with an electrical signal whose amplitude varies linearly and periodically; then injecting the generated modulated optical signal into a semiconductor slave laser to generate an optical chirp signal; and finally optically filtering the optical chirp signal to retain only a single-sided first-order sideband, thereby obtaining the linear frequency modulated optical signal.

[0020] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0021] (1) Based on the photon parameter mapping architecture, the present invention maps the frequency, phase and time information of high-speed radar signals to the intensity and time information of low-speed output pulse signals, which greatly reduces the subsequent data processing volume. In addition, the present invention adopts a layered radar signal recognition algorithm to first identify the modulation type and then identify the radar signal state based on the identified modulation type time series, thereby achieving accurate and rapid recognition of the modulation type and state.

[0022] (2) The present invention further generates an optical high-speed chirp signal based on an optical injection architecture, without the need for an external high-speed frequency sweep source, and the optical high-speed chirp signal parameters can be reconstructed simply by adjusting the low-speed electrical drive signal parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1This is a schematic diagram of the structural principle of a preferred embodiment of a radar signal recognition device based on photon parameter mapping of the present invention;

[0024] Figure 2 Obtain results for the time, frequency and phase parameters of radar signals. DETAILED DESCRIPTION

[0025] To address the shortcomings of existing technologies, the solution of the present invention is to map the frequency, phase and time information of high-speed radar signals to the intensity and time information of low-speed output pulse signals through a photon parameter mapping architecture, and then combine it with a layered radar signal recognition algorithm to achieve accurate and rapid recognition of the radar signal modulation type and state.

[0026] The radar signal recognition method based on photon parameter mapping proposed in the present invention includes the following steps:

[0027] S1, generate two optical carriers of the same source;

[0028] S2. Processing the first optical carrier into a linear frequency modulated optical signal; intensity modulating the second optical carrier with the radar signal to be measured, and introducing a time delay difference between the two orthogonal polarization components of the generated modulated optical signal to obtain a time domain differential interference optical signal;

[0029] S3. Combining the linear frequency modulated optical signal with the time-domain differential interference optical signal, performing balanced photoelectric detection, and electrically filtering the resulting electrical signal with an electrical filter and then performing envelope detection, thereby obtaining an electrical signal having time-domain pulse characteristics; the electrical filter is a narrowband filter having a passband center frequency within the frequency difference between the linear frequency modulated optical signal and the radar signal to be measured;

[0030] S4. Convert the electrical signal with time-domain pulse characteristics into a digital signal, and input the digital signal into a modulation type recognition module to identify the modulation type of the radar signal to be measured; then input the obtained modulation type time series into a state recognition module to identify the state of the radar signal to be measured; the modulation type recognition module and the state recognition module are both pre-trained deep learning models.

[0031] The radar signal recognition device based on photon parameter mapping proposed in the present invention includes:

[0032] An optical carrier generation module, used to generate two optical carriers of the same source;

[0033] A linear frequency modulated optical signal generation module, configured to process the first optical carrier into a linear frequency modulated optical signal;

[0034] A time-domain differential interference optical signal generation module is used to modulate the intensity of the second optical carrier with the radar signal to be measured, and introduce a time delay difference between the two orthogonal polarization components of the generated modulated optical signal to obtain a time-domain differential interference optical signal;

[0035] A photoelectric detection and filtering module is used to combine the linear frequency modulated optical signal with the time domain differential interference optical signal, perform balanced photoelectric detection, and use an electrical filter to electrically filter the resulting electrical signal and then perform envelope detection to obtain an electrical signal with time domain pulse characteristics; the electrical filter is a narrowband filter with a passband center frequency within the frequency difference range between the linear frequency modulated optical signal and the radar signal to be measured;

[0036] The digital processing module is used to convert the electrical signal with time-domain pulse characteristics into a digital signal, and input the digital signal into a modulation type recognition module to identify the modulation type of the radar signal to be measured; and then input the obtained modulation type time series into a state recognition module to identify the state of the radar signal to be measured; the modulation type recognition module and the state recognition module are both pre-trained deep learning models.

[0037] Among them, the generation of linear frequency modulated optical signals can be achieved through heterodyne modulation, that is, a broadband electrical linear frequency modulated signal is directly modulated on an optical carrier through an electro-optical modulator to generate a linear frequency modulated optical signal. However, this method requires the use of an external high-sampling rate signal generator, which makes the system complex and has poor real-time performance. Alternatively, it can be generated through dispersion stretching, that is, a narrowband electrical linear frequency modulated signal is modulated on an optical carrier and then connected to a dispersion module for time-frequency domain stretching, thereby increasing the generated signal bandwidth. However, the signal bandwidth of this method is limited by the dispersion amount, and it cannot be arbitrarily reconstructed and adjusted, resulting in poor reconfigurability. To this end, the present invention further generates an optical high-speed chirped signal based on an optical injection architecture, eliminating the need for an external high-speed frequency sweep source. The optical high-speed chirped signal parameters can be reconstructed simply by adjusting the parameters of a low-speed electrical drive signal. Specifically, the following method is used to process a first optical carrier into a linear frequency modulated optical signal: intensity modulating the first optical carrier using an electrical signal with a periodic linear amplitude variation; then injecting the generated modulated optical signal into a semiconductor slave laser to generate an optical chirped signal; and finally, optically filtering the optical chirped signal to retain only a single-sided first-order sideband, thereby obtaining the linear frequency modulated optical signal.

[0038] To facilitate public understanding, the technical solution of the present invention is described in detail below through a preferred embodiment with reference to the accompanying drawings:

[0039] like Figure 1As shown, the radar signal recognition device of this embodiment includes: 1 master laser, 2 couplers, 2 electro-optical modulators, 1 waveform generator, 1 slave laser, 1 optical filter, 1 polarization demultiplexer, 1 optical fiber, 1 polarization multiplexer, 1 balanced photodetector, 1 electrical filter, 1 envelope detector, 1 low-speed analog-to-digital converter, a modulation type recognition module, and a state recognition module.

[0040] like Figure 1 As shown, the optical carrier generated by the main laser is divided into two paths by the coupler.

[0041] In the upper branch, a waveform generator generates an electrical signal with a periodic, linearly varying amplitude to intensity modulate the optical carrier of the upper branch. This modulated optical signal is then injected into the slave laser via a circulator, generating an optical chirp signal controlled by a low-speed electrical signal. The characteristics of this optical chirp signal, such as the frequency modulation rate, can be flexibly adjusted by the loaded electrical control signal. An optical filter is then used to remove the optical carrier and higher-order sidebands from the optical chirp signal, retaining only the single-sided first-order optical sideband, resulting in a linear frequency modulated optical signal.

[0042] In the lower branch, the received radar signal to be measured is intensity modulated on the optical carrier of the lower branch by an electro-optical modulator. The generated modulated optical signal is input into a photon-assisted interference structure. The photon-assisted interference structure consists of a polarization demultiplexer, optical fiber, and a polarization multiplexer. It is used to introduce a specific time delay to one of the polarization components of the modulated optical signal, thereby forming a time delay difference with the other orthogonal polarization component, and ultimately generating a time-domain differential interference optical signal.

[0043] The upper and lower optical signals are combined in an optical coupler and then fed into a balanced photodetector for coherent detection. The electrical signal output by the balanced photodetector is further processed by an electrical filter and envelope detector to obtain an electrical signal with time-domain pulse characteristics. The electrical filter is a narrowband filter with a passband center frequency within the frequency difference between the linear frequency modulated optical signal and the radar signal to be measured. This ensures that the beat frequency signal generated by heterodyne mixing falls within the filter's passband, thus enabling effective signal detection.

[0044] The electrical signal with time-domain pulse characteristics is converted to the digital domain through digital sampling. The modulation type of the radar signal to be tested is first identified by the modulation type recognition module. The obtained modulation type time series is then input into the state recognition module to identify the state of the radar signal to be tested. The modulation type recognition module and the state recognition module are both pre-trained deep learning models.

[0045] The implementation principle of the device is as follows:

[0046] The optical carrier signal generated by the master laser can be expressed as

[0047] E c (t) = E0exp(j2πf0t) (1)

[0048] Among them, E0 and f0 represent the laser amplitude and frequency. In the upper branch, taking the triangle wave as an example, the low-speed electrical drive signal generated by the waveform generator is expressed as:

[0049]

[0050] Among them, V max , K and T represent the maximum amplitude, slope and period of the driving signal respectively. The driving signal is modulated on the optical carrier and filtered through an optical filter to obtain an optical chirp signal, which can be expressed as

[0051]

[0052] Among them, f max and k represent the maximum frequency and chirp rate of the optical chirp signal, respectively. In the lower branch, the received radar signal can be expressed as

[0053]

[0054] Among them, V m 、f m and Represent the amplitude, frequency and phase of the radar signal respectively. The radar signal is modulated on the optical carrier to obtain the optical radar signal, which can be expressed as

[0055]

[0056] Among them, J n (·) represents the nth order Bessel function. Connect the light-borne radar signal to the photon-assisted interference module to obtain the output signal

[0057]

[0058] Where t0 represents the delay difference introduced by the optical fiber, and α represents the amplitude ratio of the signal between time t+t0 and t. After coupling the optical radar signal and the optical chirp signal, the output signal is obtained by photoelectric detection, which can be expressed as

[0059]

[0060] Where η is the responsivity of the balanced detector. The output signal is connected to the electrical filter and envelope detector to obtain a low-speed output pulse signal, which can be expressed as:

[0061]

[0062] Among them, δ(t) represents the time domain envelope response function of the pulse, f IF is the frequency of the electric filter. It can be seen that two pulses will appear in one measurement period T, and the time interval between the pulses is 2(f max +f IF -f m ) / k, is related to the radar signal input frequency, while the amplitudes of the two pulses are related to the radar signal phase. Therefore, by periodically measuring the amplitude and time domain parameters of the output pulse signal, the time, phase, and frequency parameters of the radar signal can be obtained.

[0063] Because modern multifunction radar systems operate in complex and ever-changing states, encompassing a variety of modulation formats, parameter combinations, and state behaviors, direct "one-step" recognition of the entire radar signal faces challenges such as high parameter dimensionality and a large time span. Different radar states contain multiple modulated "radar words," resulting in complex timing sequences. Furthermore, robustness is limited, with accuracy significantly reduced in the presence of low signal-to-noise ratios and multi-state interference.

[0064] Therefore, the introduction of a hierarchical recognition architecture can significantly improve recognition accuracy. This architecture divides received radar signals into three levels: radar pulses, radar words (modulation type), and radar phrases (states). Radar pulses are basic units described by pulse description words (PDWs). PDWs consist of multiple key parameters, such as repetition rate, frequency, and pulse width. Using convolutional neural networks (CNNs) or residual neural networks (ResNets), multiple radar pulses can be identified as a radar word (modulation type) based on the changing characteristics of radar signal parameters. Finally, the identified radar word sequence is modeled as a time series. Using a gated recurrent unit (GRU) network or a long short-term memory (LSTM) network, the radar phrase (state) corresponding to the radar word time series is identified based on context and prior behavioral patterns.

[0065] During training, the pulse layer data and state layer data are first labeled to construct supervised samples and sequence classification data sets; then the aforementioned algorithm network is used for supervised training, and learning rate warm-up, AdamW optimizer, dropout and other techniques are combined to prevent overfitting; finally, the algorithm network is feedback-optimized based on the recognition accuracy of the radar modulation type and state.

[0066] To verify the system's ability to acquire radar signal parameters, the radar signal was set to a phase-shifted (BPSK) signal with a frequency range of 19-31 GHz and a signal-to-noise ratio (SNR) range of -20-11 dB. The modulation pattern used was the Barker code (1 1 1 01) with a symbol rate of 0.25 MHz. Figure 2 (a1), (a2) and (a3) ​​in the figure show the time-frequency signal diagrams of the original input under different signal-to-noise ratio (SNR) conditions; Figure 2(b1), (b2) and (b3) in the figure respectively give the corresponding output electrical signal pulses; Figure 2 (c1), (c2) and (c3) in the figure give the phase distribution images of the original input and output signals respectively; Figure 2 (d1), (d2) and (d3) in the figure respectively give the time-frequency signal diagrams of the corresponding output. The experimental results show that the time intervals between the two pulses are 1.85, 0.40 and 1.00 microseconds, respectively, and the corresponding carrier frequencies are 20, 28.6 and 25 GHz, which are consistent with the input signal frequency, verifying the accuracy of the system frequency parameter acquisition. According to formula (8), the phase change information of the radar signal can be obtained by inverting the amplitude change of the output electrical signal pulse. Taking the BPSK signal as an example, when a 180° phase reversal occurs, the corresponding normalized electrical signal amplitude is about 0.6; when the phase remains stable, the normalized amplitude is close to 1. Figure 2 Taking (b1) as an example, for each 20μs input signal, the corresponding output pulse is divided into five segments, and the amplitude of each output pulse reflects the phase relationship between the current moment and the previous 4μs. Figure 2 The normalized pulse amplitudes of the five segments in (b1) are 0.59, 0.59, 0.89, 0.89, and 0.93, respectively. According to formula (8), it can be determined that the phase of segments 1 and 2 has changed. Assuming the initial phase code of the radar signal is 1, its phase code can be restored to (01111), which is consistent with the actual input signal, verifying the phase determination accuracy of the system.

[0067] Compared with the traditional digital radar modulation recognition system, the system of the present invention has achieved significant optimization in signal acquisition and processing performance for broadband radar signal sequences with a frequency range of 19-29GHz and a duration of 8 milliseconds. In the digital acquisition and recognition solution based on electronic technology, a high-speed analog-to-digital converter of 58GSa / s is required to sample the signal, with a computational complexity of approximately 26427GFLOPs and an overall processing time of approximately 80 milliseconds. The radar signal recognition device based on photon parameter mapping proposed in the present invention, combined with frequency mapping and deep neural network processing mechanism, effectively reduces the sampling rate to 150MSa / s, and the corresponding computational complexity is greatly reduced to 64.98GFLOPs, and the processing delay is shortened to 0.2 milliseconds, which greatly reduces system resource consumption and response time.

[0068] In summary, the system of the present invention is superior to existing traditional methods in terms of sampling rate requirements, computational complexity and processing delay, and is particularly suitable for application scenarios with extremely high processing performance requirements, such as broadband radar modulation recognition, cognitive electronic countermeasures, and dynamic spectrum monitoring.

Claims

1. A radar signal recognition method based on photon parameter mapping, characterized in that: The following steps are involved: S1, generate two optical carriers of the same source; S2. Processing the first optical carrier into a linear frequency modulated optical signal; intensity modulating the second optical carrier with the radar signal to be measured, and introducing a time delay difference between the two orthogonal polarization components of the generated modulated optical signal to obtain a time domain differential interference optical signal; S3. Combining the linear frequency modulated optical signal with the time-domain differential interference optical signal, performing balanced photoelectric detection, and electrically filtering the resulting electrical signal with an electrical filter and then performing envelope detection, thereby obtaining an electrical signal having time-domain pulse characteristics; the electrical filter is a narrowband filter having a passband center frequency within the frequency difference between the linear frequency modulated optical signal and the radar signal to be measured; S4. Convert the electrical signal with time-domain pulse characteristics into a digital signal, and input the digital signal into a modulation type recognition module to identify the modulation type of the radar signal to be measured; then input the obtained modulation type time series into a state recognition module to identify the state of the radar signal to be measured; the modulation type recognition module and the state recognition module are both pre-trained deep learning models.

2. The radar signal recognition method based on photon parameter mapping according to claim 1, characterized in that: The first optical carrier is processed into a linear frequency modulated optical signal using the following method: the first optical carrier is intensity modulated using an electrical signal with a periodic linear amplitude change; the generated modulated optical signal is then injected into a semiconductor slave laser to generate an optical chirp signal; and finally, the optical chirp signal is optically filtered to retain only a single-sided first-order sideband, thereby obtaining the linear frequency modulated optical signal.

3. A radar signal recognition device based on photon parameter mapping, characterized in that: include: An optical carrier generation module, used to generate two optical carriers of the same source; A linear frequency modulated optical signal generation module, configured to process the first optical carrier into a linear frequency modulated optical signal; A time-domain differential interference optical signal generation module is used to modulate the intensity of the second optical carrier with the radar signal to be measured, and introduce a time delay difference between the two orthogonal polarization components of the generated modulated optical signal to obtain a time-domain differential interference optical signal; A photoelectric detection and filtering module is used to combine the linear frequency modulated optical signal with the time domain differential interference optical signal, perform balanced photoelectric detection, and use an electrical filter to electrically filter the resulting electrical signal and then perform envelope detection to obtain an electrical signal with time domain pulse characteristics; the electrical filter is a narrowband filter with a passband center frequency within the frequency difference range between the linear frequency modulated optical signal and the radar signal to be measured; The digital processing module is used to convert the electrical signal with time-domain pulse characteristics into a digital signal, and input the digital signal into a modulation type recognition module to identify the modulation type of the radar signal to be measured; and then input the obtained modulation type time series into a state recognition module to identify the state of the radar signal to be measured; the modulation type recognition module and the state recognition module are both pre-trained deep learning models.

4. The radar signal recognition device based on photon parameter mapping according to claim 3, characterized in that: The first optical carrier is processed into a linear frequency modulated optical signal using the following method: the first optical carrier is intensity modulated using an electrical signal with a periodic linear amplitude change; the generated modulated optical signal is then injected into a semiconductor slave laser to generate an optical chirp signal; and finally, the optical chirp signal is optically filtered to retain only a single-sided first-order sideband, thereby obtaining the linear frequency modulated optical signal.

Images