A method for intra-pulse modulation type decision by decomposing and reconstructing a pulse signal

Abstract

This application provides a method for determining the intra-pulse modulation type of a decomposed and reconstructed pulse signal, including: determining the positions of the pulse rising and falling edges; obtaining pulse overlap points based on prior information; and comprehensively identifying the intra-pulse modulation type of the overlap points based on the distortion of the overlap point envelope and the analysis results of each sub-signal. This application can achieve correct determination of the intra-pulse modulation type of a pulse signal after signal decomposition and sparse reconstruction.

Description

Technical Field

[0001] This application relates to the field of radar signal processing technology, and in particular to a method for determining the intra-pulse modulation type of a decomposed and reconstructed pulse signal. Background Technology

[0002] As the electromagnetic environment becomes increasingly complex, electronic reconnaissance systems often receive signals that are superimposed in the spatial, time, and frequency domains, with overlaps between them. In the case of a single channel (or when multiple signals enter from the main lobe and cannot be distinguished in the spatial domain), the constituent signals overlap in the time, frequency, and time-frequency domains. Signal separation is often referred to as single-channel signal separation.

[0003] A hybrid signal separation method based on signal decomposition and sparse reconstruction, under the condition of known number of signal sources and instantaneous frequency, establishes a sparse observation model of the received signal by performing Fourier decomposition on each component signal. This transforms the signal separation problem into a multi-signal sparse reconstruction problem, realizing the separation of radar, communication, and other signals in complex electromagnetic environments, and providing a new technical approach for hybrid multi-signal separation under single-channel conditions. However, this method suffers from problems such as signal envelope and phase distortion after separation, which significantly affects the intra-pulse analysis and intra-pulse modulation type determination of the separated signal.

[0004] For signals separated by a hybrid signal separation method based on signal decomposition and sparse reconstruction, current intra-pulse modulation type decision algorithms struggle to accurately determine their separation. Summary of the Invention

[0005] This application provides a method for determining the intra-pulse modulation type of a decomposed and reconstructed pulse signal, which can be used to solve the technical problem of being unable to accurately determine the intra-pulse analysis and intra-pulse modulation type of the separated signal.

[0006] This application provides a method for determining the intra-pulse modulation type of a decomposed and reconstructed pulse signal, the method comprising:

[0007] Step 1: Determine the positions of the rising and falling edges of the pulse;

[0008] Step 2: Obtain the pulse overlap point based on prior information. Based on the distortion of the overlap point envelope, set the sampling points near the overlap point to zero and divide the pulse into several sub-signals accordingly.

[0009] Step 3: Perform intra-pulse analysis on each segment of the signal in sequence;

[0010] Step four: Based on the analysis results of each sub-signal, perform a comprehensive identification of the intra-pulse modulation type.

[0011] Optionally, determine the positions of the pulse rising and falling edges, including:

[0012] Calculate the envelope a(n) of the pulse digital sampling sequence x(n), and calculate the average value of the first 20 sampling points of the pulse envelope sequence a(n), denoted as a. min ;

[0013] Find the position m of the maximum value of a(n), and calculate the average of the 20 sampling points before and after the position of the maximum value, denoted as a. max ;

[0014] Starting from a(n), find values ​​greater than or equal to... The first point, corresponding to the position of the rising edge of the pulse, is denoted as k1;

[0015] Starting from the endpoint of a(n), find values ​​greater than or equal to... The first point, corresponding to the position of the falling edge of the pulse, is denoted as k2;

[0016] Extract a(k1) to a(k2) as the new pulse to be analyzed.

[0017] Optional, the rules for setting to zero are as follows:

[0018] Zeroing is performed by traversing outwards from the overlap point. If the difference between the envelope value and the envelope mean at the sampling point is within ±30%, the traversal and zeroing stops. The number of zeroing points is at least 10.

[0019] Optionally, intra-pulse analysis can be performed on each segment of the signal sequentially, including:

[0020] Step 31: Perform FFT peak search and bandwidth estimation;

[0021] FFT spectrum estimation is performed on the sub-signal to estimate the center frequency and bandwidth of the signal.

[0022] The upper and lower limits of the signal frequency are estimated by searching the upper and lower boundary edges, and the center frequency f0 and bandwidth B of the signal are estimated by averaging.

[0023] Step 32, Dynamic Filter Generation

[0024] Based on the center frequency f0 and bandwidth B, construct an FIR filter with a bandwidth of 1.5B;

[0025] Step 33, Fixed intermediate frequency conversion and filtering

[0026] The signal is converted to a fixed intermediate frequency using digital frequency conversion methods.

[0027] Step 34, Hilbert filtering

[0028] The Hilbert filter method is used to orthogonally transform the preprocessed fixed intermediate frequency signal;

[0029] Step 35, Phase Calculation and Defuzzification

[0030] The signal after Hilbert transform is determined by the following method:

[0031] r(n)=s(n)+w(n)=Aexp{jφ(n)}+w(n),n=1,…N

[0032] Where A is the carrier amplitude, φ(n) is the phase, s(n) is the true signal, and w(n) is the variance equal to σ. 2 The additional Gaussian white noise, where N is the number of samples;

[0033] The measured phase is obtained from the complex signal through arctangent:

[0034]

[0035] Im[r(n)] represents the imaginary part of the signal after the Hilbert transform, and Re[r(n)] represents the real part of the signal after the Hilbert transform;

[0036] Since the measured phase ranges from [0, 2π), the measured phase φ(n) is the true phase φ. T (n) Phase value modulo 2π:

[0037] φ(n) = mod(φ) T (n),2π)

[0038] Where mod represents the modulo operation, the above equation is equivalent to:

[0039] φ(n)=φ T (n)-k n 2π

[0040] Where kn is the period number of the nth sample point, and the phase difference between two adjacent sample points is:

[0041] Δφ(n)=Δφ T (n)-(k n -k n-1 2π

[0042] Under the condition that the Nyquist sampling theorem is satisfied, if φ T (n) and φ T If (n-1) are in adjacent periods, then Δφ(n) < 0. The true phase of the signal is obtained by deblurring step by step based on the measured phase difference between adjacent sample points. The true phase value of the nth sample point is:

[0043]

[0044] When the phase is contaminated by noise, the phase difference between adjacent sample points is:

[0045] Δφ(n)=Δφ T (n)-(k n -k n-1 )2π+Δε(n)

[0046] In the formula, Δε(n) represents phase noise;

[0047] When φ T (n) and φ T (n-1) are within the same period:

[0048] Δφ(n)=Δφ T (n)+Δε(n)

[0049] When the signal-to-noise ratio is low, the sampled signal is up-converted to approximately half of the signal sampling rate fs, and the phase is expanded using the following formula to obtain the accurate true phase φ. T (n),

[0050]

[0051] Then, the true phase of the signal before frequency conversion is obtained based on the frequency of the up-conversion.

[0052] Step 36, High-order polynomial phase fitting

[0053] Step 37, Intra-pulse modulation identification based on decision tree

[0054] For the fitted phase curve Feature extraction is performed to construct an intra-pulse modulation recognition decision tree to identify the intra-pulse modulation type.

[0055] Optionally, based on the analysis results of each sub-signal, a comprehensive identification of the intra-pulse modulation type can be performed, including:

[0056] If the sub-signal is identified as an LFM or NLFM signal, then a quadratic linear fit is performed based on the start and end frequencies and the length of the zero point of each sub-signal segment. If the result of the quadratic linear fit is a straight line, and the correlation between the straight line and the sub-signal segment in the overlapping segment of each sub-signal segment is greater than 0.9, then the signal is an LFM signal; otherwise, the signal is an NLFM signal.

[0057] If the sub-signal is identified as a phase-coded signal or a regular pulse signal, then if any sub-signal is QPSK, then the signal is QPSK; if there is a sub-signal that is BPSK and no sub-signal that is QPSK, then the signal is BPSK.

[0058] If the sub-signal is identified as an FSK signal and a regular pulse signal, then the signal is an FSK signal;

[0059] If all sub-signals are regular pulses, then the signal estimates the carrier frequency of each sub-signal. If the carrier frequency difference between any two sub-signals is less than 2MHz, then the signal is a regular pulse signal; otherwise, the signal is an FSK signal.

[0060] If the intrapulse type identified by the sub-signal is not among the above conditions, the signal is classified as a composite modulation signal.

[0061] Pulse signals that have undergone signal decomposition and sparse reconstruction exhibit significant distortion in the envelope and phase relationship at their time-frequency overlap positions, making it difficult for conventional intra-pulse modulation type identification algorithms to correctly identify their modulation type. This application first analyzes the pulse envelope to accurately determine the pulse's leading and trailing edges. Then, through judgment, filtering, and processing of the distorted pulse envelope region, the processed pulse signal data is used to determine the intra-pulse modulation type, enabling accurate identification of the intra-pulse modulation type for pulse signals that have undergone signal decomposition and sparse reconstruction. Simulation results show that it can effectively adapt to the identification of intra-pulse modulation types such as conventional pulse signals, LFM, BPSK, and QPSK. Attached Figure Description

[0062] Figure 1 This application provides a flowchart of a method for determining the intra-pulse modulation type of a decomposed and reconstructed pulse signal, as described in an embodiment of the present application. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0064] The embodiments of this application will now be described in conjunction with the accompanying drawings.

[0065] This application provides a method for determining the intra-pulse modulation type of a decomposed and reconstructed pulse signal, including:

[0066] Step 1: Determine the positions of the rising and falling edges of the pulse.

[0067] Calculate the envelope a(n) of the pulse digital sampling sequence x(n), and calculate the average value of the first 20 sampling points of the pulse envelope sequence a(n), denoted as a. min ;

[0068] Find the position m of the maximum value of a(n), and calculate the average of the 20 sampling points before and after the position of the maximum value, denoted as a. max ;

[0069] Starting from a(n), find values ​​greater than or equal to... The first point, corresponding to the position of the rising edge of the pulse, is denoted as k1;

[0070] Starting from the endpoint of a(n), find values ​​greater than or equal to... The first point, corresponding to the position of the falling edge of the pulse, is denoted as k2;

[0071] Extract a(k1) to a(k2) as the new pulse to be analyzed.

[0072] Step 2: Obtain the pulse overlap point based on prior information. Based on the distortion of the overlap point envelope, set the sampling points near the overlap point to zero, and divide the pulse into several sub-signals accordingly.

[0073] The rules for setting to zero are as follows:

[0074] Zeroing is performed by traversing outwards from the overlap point. If the difference between the envelope value and the envelope mean at the sampling point is within ±30%, the traversal and zeroing stops. The number of zeroing points is at least 10.

[0075] Step 3: Perform intra-pulse analysis on each segment of the signal in sequence.

[0076] Specifically, in step 31, FFT peak search and bandwidth estimation are performed;

[0077] FFT spectrum estimation is performed on the sub-signal to estimate the center frequency and bandwidth of the signal.

[0078] The upper and lower limits of the signal frequency are estimated by searching the upper and lower boundary edges, and the center frequency f0 and bandwidth B of the signal are estimated by averaging.

[0079] Step 32, Dynamic Filter Generation

[0080] Based on the center frequency f0 and bandwidth B, an FIR filter with a bandwidth of 1.5B is constructed; the FIR filter is a linear phase filter and does not affect the phase of the original signal.

[0081] Step 33, Fixed intermediate frequency conversion and filtering

[0082] The signal is converted to a fixed intermediate frequency using digital frequency conversion methods.

[0083] The purpose of fixed intermediate frequency (IF) conversion is to convert the signal to a fixed IF frequency for easier post-processing. The conversion uses a digital conversion method, which avoids unnecessary intermodulation and therefore offers better performance. After conversion, additional frequency components are generated. These are digitally filtered using the filter constructed in step 2), retaining only the fixed IF frequency components.

[0084] Step 34, Hilbert filtering

[0085] The Hilbert filter method is used to orthogonally transform the preprocessed fixed intermediate frequency signal; there are two methods for orthogonal signal transformation: the Hilbert filter method and the FFT transform method. This invention uses the FFT transform method for calculation, which has a fast calculation speed.

[0086] Step 35, Phase Calculation and Defuzzification

[0087] The signal after Hilbert transform is determined by the following method:

[0088] r(n)=s(n)+w(n)=Aexp{jφ(n)}+w(n),n=1,…N

[0089] Where A is the carrier amplitude, φ(n) is the phase, s(n) is the true signal, and w(n) is the variance equal to σ. 2 The additional Gaussian white noise, where N is the number of samples;

[0090] The measured phase is obtained from the complex signal through arctangent:

[0091]

[0092] Im[r(n)] represents the imaginary part of the signal after the Hilbert transform, and Re[r(n)] represents the real part of the signal after the Hilbert transform;

[0093] Since the measured phase ranges from [0, 2π), the measured phase φ(n) is the true phase φ. T (n) Phase value modulo 2π:

[0094] φ(n) = mod(φ) T (n),2π)

[0095] Where mod represents the modulo operation, the phase expansion algorithm obtains the true phase φ by defuzzifying the measured phase φ(n). T (n). The equivalent of the above equation is:

[0096] φ(n)=φ T (n)-k n 2π

[0097] Where kn is the period number of the nth sample point, and the phase difference between two adjacent sample points is:

[0098] Δφ(n)=Δφ T (n)-(k n -k n-1 2π

[0099] Under the condition that the Nyquist sampling theorem is satisfied, if φ T (n) and φ T If (n-1) are in adjacent periods, then Δφ(n) < 0. The true phase of the signal is obtained by deblurring step by step based on the measured phase difference between adjacent sample points. The true phase value of the nth sample point is:

[0100]

[0101] When the phase is contaminated by noise, the phase difference between adjacent sample points is:

[0102] Δφ(n)=Δφ T (n)-(k n -k n-1 )2π+Δε(n)

[0103] In the formula, Δε(n) represents phase noise;

[0104] When φ T (n) and φ T (n-1) are within the same period:

[0105] Δφ(n)=Δφ T (n)+Δε(n)

[0106] When Δε(n) <-Δφ T When (n), the φ obtained by defuzzification T Errors may occur with (n). To reduce the probability of errors, then Δφ T (n) should be as large as possible because Δφ T (n)∈(0,π], therefore when Δφ T When (n) = π, the phase expansion performance is optimal;

[0107] When the signal-to-noise ratio is low, the sampled signal is up-converted to approximately half of the signal sampling rate fs, and the phase is expanded using the following formula to obtain the accurate true phase φ. T (n),

[0108]

[0109] Then, the true phase of the signal before frequency conversion is obtained based on the frequency of the up-conversion.

[0110] Step 36, High-order polynomial phase fitting

[0111] To reduce the impact of noise, it is necessary to adjust φ T (n) Perform high-order polynomial fitting; the higher the order used in the fitting, the greater the computational cost and the higher the fitting accuracy, especially for fitting complex modulation types. Therefore, an empirical value of 19th order or higher is used for fitting. The fitted phase curve is φ P (n).

[0112] Step 37, Intra-pulse modulation identification based on decision tree

[0113] For the fitted phase curve Feature extraction is performed to construct an intra-pulse modulation recognition decision tree to identify the intra-pulse modulation type. The identified types include conventional pulse, linear frequency modulation (LFM), nonlinear frequency modulation (NLFM), binary phase-coded (BPSK), quadrature phase-coded (QPSK), and frequency keying (FSK).

[0114] Step four: Based on the analysis results of each sub-signal, perform a comprehensive identification of the intra-pulse modulation type.

[0115] Specifically, if a sub-signal is identified as an LFM or NLFM signal, a quadratic linear fit is performed based on the start and end frequencies and the length of the zero point of each sub-signal segment. If the result of the quadratic linear fit is a straight line, and the correlation between the straight line and the sub-signal segment in the overlapping segment of each sub-signal segment is greater than 0.9, then the signal is an LFM signal; otherwise, the signal is an NLFM signal.

[0116] If the sub-signal is identified as a phase-coded signal or a regular pulse signal, then if any sub-signal is QPSK, then the signal is QPSK; if there is a sub-signal that is BPSK and no sub-signal that is QPSK, then the signal is BPSK.

[0117] If the sub-signal is identified as an FSK signal and a regular pulse signal, then the signal is an FSK signal;

[0118] If all sub-signals are regular pulses, then the signal estimates the carrier frequency of each sub-signal. If the carrier frequency difference between any two sub-signals is less than 2MHz, then the signal is a regular pulse signal; otherwise, the signal is an FSK signal.

[0119] If the intrapulse type identified by the sub-signal is not among the above conditions, the signal is classified as a composite modulation signal.

[0120] Pulse signals that have undergone signal decomposition and sparse reconstruction exhibit significant distortion in the envelope and phase relationship at their time-frequency overlap positions, making it difficult for conventional intra-pulse modulation type identification algorithms to correctly identify their modulation type. This application first analyzes the pulse envelope to accurately determine the pulse's leading and trailing edges. Then, through judgment, filtering, and processing of the distorted pulse envelope region, the processed pulse signal data is used to determine the intra-pulse modulation type, enabling accurate identification of the intra-pulse modulation type for pulse signals that have undergone signal decomposition and sparse reconstruction. Simulation results show that it can effectively adapt to the identification of intra-pulse modulation types such as conventional pulse signals, LFM, BPSK, and QPSK.

[0121] The embodiments described above do not constitute a limitation on the scope of protection of this application.

Claims

1. A method for determining the intra-pulse modulation type of a decomposed and reconstructed pulse signal, characterized in that, The method includes: Step 1: Determine the positions of the rising and falling edges of the pulse; Step 2: Obtain the pulse overlap point based on prior information. Based on the distortion of the overlap point envelope, set the sampling points near the overlap point to zero and divide the pulse into several sub-signals accordingly. Step 3: Perform intra-pulse analysis on each segment of the signal in sequence; Step four: Based on the analysis results of each sub-signal, perform a comprehensive identification of the intra-pulse modulation type; Intra-pulse analysis is performed sequentially on each sub-signal, including: Step 31: Perform FFT peak search and bandwidth estimation; Perform FFT spectrum estimation on the sub-signal to estimate the center frequency and bandwidth of the signal; The upper and lower limits of the signal frequency are estimated by searching the upper and lower boundary edges, and the center frequency f0 and bandwidth B of the signal are estimated by averaging. Step 32, Dynamic Filter Generation Based on the center frequency f0 and bandwidth B, construct an FIR filter with a bandwidth of 1.5B; Step 33, Fixed intermediate frequency conversion and filtering The signal is converted to a fixed intermediate frequency using digital frequency conversion methods. Step 34, Hilbert filtering The Hilbert filter method is used to orthogonally transform the preprocessed fixed intermediate frequency signal; Step 35, Phase Calculation and Defuzzification The signal after Hilbert transform is determined by the following method: r(n)=s(n)+w(n)=Aexp{jφ(n)}+w(n),n=1,…N Where A is the carrier amplitude, φ(n) is the phase, s(n) is the true signal, and w(n) is the variance equal to σ. 2 The additional Gaussian white noise, where N is the number of samples; The measured phase is obtained from the complex signal through arctangent: Im[r(n)] represents the imaginary part of the signal after the Hilbert transform, and Re[r(n)] represents the real part of the signal after the Hilbert transform; Since the measured phase ranges from [0, 2π), the measured phase φ(n) is the true phase φ. T (n) Phase value modulo 2π: φ(n)=mod(φ T (n),2π) Where mod represents the modulo operation, the above equation is equivalent to: φ(n)=φ T (n)-k n 2p Where kn is the period number of the nth sample point, and the phase difference between two adjacent sample points is: Δφ(n)=Δφ T (n)-(k n -k n-1 )2π Under the condition that the Nyquist sampling theorem is satisfied, if φ T (n) and φ T If (n-1) are in adjacent periods, then Δφ(n) < 0. The true phase of the signal is obtained by deblurring step by step based on the measured phase difference between adjacent sample points. The true phase value of the nth sample point is: When the phase is contaminated by noise, the phase difference between adjacent sample points is: Δφ(n)=Δφ T (n)-(k n -k n-1 )2π+Δε(n) In the formula, Δε(n) represents phase noise; When φ T (n) and φ T (n-1) are within the same period: Δφ(n)=Δφ T (n)+No(n) When the signal-to-noise ratio is low, the sampled signal is up-converted to approximately half of the signal sampling rate fs, and the phase is expanded using the following formula to obtain the accurate true phase φ. T (n), Then, the true phase of the signal before frequency conversion is obtained based on the frequency of the up-conversion. Step 36, high-order polynomial phase fitting; Step 37, Intra-pulse modulation identification based on decision tree; For the fitted phase curve Feature extraction is performed, and an intra-pulse modulation recognition decision tree is constructed to identify the intra-pulse modulation type. Based on the analysis results of each sub-signal, a comprehensive identification of the intra-pulse modulation type is performed, including: If the sub-signal is identified as an LFM or NLFM signal, then a quadratic linear fit is performed based on the start and end frequencies and the length of the zero point of each sub-signal segment. If the result of the quadratic linear fit is a straight line, and the correlation between the straight line and the sub-signal segment in the overlapping segment of each sub-signal segment is greater than 0.9, then the signal is an LFM signal; otherwise, the signal is an NLFM signal. If the sub-signal is identified as a phase-coded signal or a regular pulse signal, then if any sub-signal is QPSK, then the signal is QPSK; if there is a sub-signal that is BPSK and no sub-signal that is QPSK, then the signal is BPSK. If the sub-signal is identified as an FSK signal and a regular pulse signal, then the signal is an FSK signal; If all sub-signals are regular pulses, then the signal estimates the carrier frequency of each sub-signal. If the carrier frequency difference between any two sub-signals is less than 2MHz, then the signal is a regular pulse signal; otherwise, the signal is an FSK signal. If the intrapulse type identified by the sub-signal is not among the above conditions, the signal is classified as a composite modulation signal.

2. The method according to claim 1, characterized in that, Determine the positions of the pulse rising and falling edges, including: Calculate the envelope a(n) of the pulse digital sampling sequence x(n), and calculate the average value of the first 20 sampling points of the pulse envelope sequence a(n), denoted as a. min ; Find the position m of the maximum value of a(n), and calculate the average of the 20 sampling points before and after the position of the maximum value, denoted as a. max ; Starting from a(n), find values ​​greater than or equal to... The first point, corresponding to the position of the rising edge of the pulse, is denoted as k1; Starting from the endpoint of a(n), find values ​​greater than or equal to... The first point, corresponding to the position of the falling edge of the pulse, is denoted as k2; Extract a(k1) to a(k2) as the new pulse to be analyzed.

3. The method according to claim 1, characterized in that, The rules for setting to zero are as follows: The zeroing process is performed by traversing outwards from the overlap point. If the difference between the envelope value and the average envelope value at the sampling point is within ±30%, the traversal and zeroing process is stopped. The number of zeroing points is at least 10.

Images