Signal intra-pulse modulation pattern identification method and system

Abstract

The invention relates to the technical field of data processing, and discloses a signal intra-pulse modulation pattern recognition method and system, and the method comprises the steps: receiving a radar pulse signal; performing short-time Fourier transform on the radar pulse signal; in the time-frequency matrix, calculating an energy distribution gradient of each time point along a frequency axis; according to the energy distribution gradient of each time point, identifying gradient abrupt change points; determining a filtering bandwidth boundary based on the gradient catastrophe point by adopting a negative feedback regulation mechanism of bandwidth and signal-to-noise ratio; based on the filtering bandwidth boundary, performing bandwidth filtering on the radar pulse signal; calculating an instantaneous frequency, a first-order differential frequency and a second-order differential frequency of the filtered signal through Hilbert transform, and forming a feature vector; predicting a cluster number range through a long and short term memory neural network according to the feature vectors; according to the feature vector, through a K-Means clustering algorithm, carrying out clustering analysis in a clustering number range; and outputting a modulation style label according to a clustering analysis result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a signal intra-pulse modulation pattern recognition method and system. BACKGROUND

[0002] In the field of electronic signal intra-pulse modulation feature extraction, the complete processing flow usually includes key links such as signal reception, preprocessing, feature extraction and analysis. After the signal reception device captures the original signal from the complex electromagnetic environment, it needs to go through filtering, noise reduction and other preprocessing operations to remove interference and improve signal quality.

[0003] In the prior art, the preprocessing of low interception targets such as electronic signals mainly adopts a fixed bandwidth filtering method. This method filters noise of the received signal by pre-setting a fixed filtering bandwidth, and relies on bandwidth parameters set by artificial or experience. The core logic is to retain signal energy based on the pre-set bandwidth range and filter noise interference outside the bandwidth.

[0004] However, in a complex battlefield electromagnetic environment, the signal signal-to-noise ratio is extremely low, and the target signal is easily submerged in noise. Since the fixed bandwidth filtering method cannot dynamically adjust the bandwidth parameters according to the actual electromagnetic environment changes, it cannot adapt to the dynamic distribution of signal time-frequency energy. When the noise frequency falls within the pre-set bandwidth, it cannot effectively filter the interference; and when the signal bandwidth exceeds the pre-set range due to environmental changes, the key features (such as instantaneous frequency) of the signal are lost, resulting in a high signal recognition failure rate under low signal-to-noise ratio, which is difficult to meet the detection needs of low interception targets. SUMMARY

[0005] To solve the above technical problems of the prior art, the present application provides a signal intra-pulse modulation pattern recognition method, comprising: S1: receiving a radar pulse signal; S2: performing short-time Fourier transform on the radar pulse signal to generate a time-frequency matrix; S3: in the time-frequency matrix, calculating the energy distribution gradient of each time point along the frequency axis; S4: identifying a gradient mutation point according to the energy distribution gradient of each time point; S5: adopting a negative feedback adjustment mechanism of bandwidth and signal-to-noise ratio, determining a filtering bandwidth boundary based on the gradient mutation point; S6: performing bandwidth filtering on the radar pulse signal based on the filtering bandwidth boundary; S7: calculating the instantaneous frequency, first-order differential frequency and second-order differential frequency of the filtered signal through Hilbert transform, and forming a feature vector; S8: predicting a clustering number range through a long short-term memory neural network according to the feature vector; S9: According to the feature vector, clustering analysis is performed in the range of the number of clusters by a K-Means clustering algorithm. S10: According to the clustering analysis result, a modulation pattern label is output.

[0006] In a second aspect of the present application, a signal intra-pulse modulation pattern recognition system is provided, comprising a processor and a memory. The memory stores programs or instructions executable on the processor, and the programs or instructions, when executed by the processor, implement the steps of the signal intra-pulse modulation pattern recognition method according to the first aspect.

[0007] The beneficial effects of the present application are embodied in that the negative feedback adjustment mechanism of bandwidth and signal-to-noise ratio can dynamically adjust the filtering bandwidth according to the time-frequency characteristics of the signal. When the signal-to-noise ratio is high, the bandwidth can be contracted to avoid unnecessary noise interference. When the signal-to-noise ratio is low, the bandwidth is automatically expanded to ensure that the key features of the signal are not lost, effectively improving the signal recognition accuracy under low signal-to-noise ratio, enhancing the robustness of signal processing, reducing signal loss due to bandwidth mismatch, improving the reliability in low interception target detection, and ultimately achieving more accurate signal recognition and modulation pattern classification. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 A flowchart of a signal intra-pulse modulation pattern recognition method provided by the present application.

[0009] Figure 2 A flowchart of a dynamic adjustment method of the window length of a sliding time window provided by the present application.

[0010] Figure 3 A flowchart of a determination method of the filtering bandwidth boundary provided by the present application.

[0011] Figure 4 A flowchart of a clustering analysis method provided by the present application.

[0012] Figure 5 A structural diagram of a signal intra-pulse modulation pattern recognition system provided by the present application. DETAILED DESCRIPTION

[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0014] Embodiment 1: Referring to Figures 1-4 The signal pulse-in modulation pattern recognition method provided by the embodiment of the present application comprises: S1: receiving a radar pulse signal.

[0015] It should be noted that this step is the basic input link of signal processing.

[0016] In a possible implementation, after S1 and before S2, the radar pulse signal is further converted into a digital signal through analog-to-digital conversion.

[0017] It should be noted that the radar pulse signal is converted into a digital signal through analog-to-digital conversion (ADC), which can adapt the signal to modern digital signal processing technology, facilitate subsequent time-frequency analysis, feature extraction and intelligent recognition, and at the same time, the digital signal has strong anti-interference ability, high processing precision, can flexibly adjust parameters and support storage and reproduction, thereby improving the accuracy and system compatibility of modulation pattern recognition.

[0018] S2: performing short-time Fourier transform on the radar pulse signal to generate a time-frequency matrix.

[0019] The short-time Fourier transform (STFT) is an analysis method for decomposing a signal into a time-frequency domain, which divides the signal into short-time segments through sliding window processing, and applies Fourier transform on each segment to obtain the frequency distribution information at each time point. STFT can effectively reveal the changes of a signal in time and frequency, and is widely used in the analysis of non-stationary signals such as speech, audio, radar signals, etc. Its main feature is to provide local time-frequency information of the signal, and is suitable for real-time monitoring and analysis of time-varying signals.

[0020] In a possible implementation, S2 specifically comprises sub-steps S201 and S202: S201: dynamically adjusting the window length of the sliding time window according to the signal carrier frequency.

[0021] The window length of the sliding time window is 512 sampling points by default.

[0022] Optionally, S201 specifically comprises S2011 to S2015: S2011: before processing each frame of signal, performing frequency spectrum analysis on the previous frame of signal through fast Fourier transform to extract the main peak frequency.

[0023] Among them, the fast Fourier transform (FFT) is a method of efficiently calculating the discrete Fourier transform (DFT), which greatly improves the calculation speed of the Fourier transform by reducing redundant calculations. The core principle of FFT is to use the divide-and-conquer method to decompose a large-scale Fourier transform problem into multiple small-scale problems, thereby improving the computational efficiency. It is widely used in the field of signal processing, such as audio analysis, image processing, communication systems, etc., for converting time-domain signals into frequency-domain signals to reveal the spectral information of the signal.

[0024] Among them, the peak frequency refers to the frequency point with the maximum energy in the signal spectrum, that is, the most prominent peak position on the spectrum graph. In practical applications, the peak frequency usually represents the center frequency or carrier frequency of the signal, which is the main frequency component of the signal.

[0025] S2012: Take the peak frequency as the current carrier frequency value, compare it with the initial carrier frequency, and calculate the carrier frequency deviation: ; Among them, represents the carrier frequency deviation, represents the peak frequency, represents the initial carrier frequency.

[0026] S2013: When the carrier frequency deviation satisfies , keep the window length of the sliding time window unchanged.

[0027] It should be noted that when the carrier frequency deviation is small, the signal frequency changes little, and keeping the current window length can reduce the complexity of adjustment and improve the processing efficiency. At this time, there is no need to adjust frequently, reducing the computational burden.

[0028] S2014: When the carrier frequency deviation satisfies , adjust the window length of the sliding time window: ; Among them, represents the adjusted window length, L represents the current window length.

[0029] It should be noted that in the case of moderate carrier frequency change (2-5MHz), the window length can adapt to the signal frequency change, and by adjusting the window length, it can ensure that the short-time Fourier transform (STFT) can always capture the key frequency components of the signal, ensuring the accuracy of time-frequency analysis.

[0030] S2015: When the carrier frequency deviation satisfies , trigger window length reset and re-execute initial value setting to ensure that the window length of the time window is adapted to the new carrier frequency.

[0031] It should be noted that when the carrier frequency changes greatly, resetting the window length can ensure the flexibility and adaptability of signal analysis. When the carrier frequency changes greatly, resetting the window length is necessary, which can prevent the fixed window from failing to adapt to the wide frequency signal, ensure that the time-frequency analysis of the signal can cover the new frequency range, and avoid loss of signal information.

[0032] S202: Calculate the frequency component of each window by short-time Fourier transform, and construct a three-dimensional time-frequency matrix containing time, frequency and energy.

[0033] The prior art STFT window is fixed, and the window length of the present application can be adaptively adjusted according to the signal carrier frequency, avoiding the problem of insufficient adaptability of the fixed window to wide and narrow frequency signals.

[0034] S3: Calculate the energy distribution gradient of each time point along the frequency axis in the time-frequency matrix.

[0035] In one possible implementation, S3 is specifically: calculate the energy distribution gradient by the energy difference of adjacent frequency points: ; Wherein, represents t the energy distribution gradient of the frequency at f , represents t the energy value at f +1, represents t the energy value at f -1.

[0036] It should be noted that using the energy difference of adjacent frequency points to calculate the gradient can quickly capture the local change characteristics of the energy in the time-frequency matrix, highlight the mutation of the signal on the frequency axis (such as the slope of the linear frequency modulation or the phase jump), and at the same time suppress the interference of gentle noise. This calculation method is simple and efficient, suitable for real-time processing, and provides accurate local change information for subsequent gradient mutation point detection and bandwidth boundary extraction.

[0037] S4: Identify the gradient mutation point according to the energy distribution gradient of each time point.

[0038] Specifically, the data point whose gradient absolute value exceeds the preset threshold is the gradient mutation point. The preset threshold is 10% of the average energy of the current frequency band.

[0039] S5: Adopt a negative feedback adjustment mechanism of bandwidth and signal-to-noise ratio, and determine the filter bandwidth boundary based on the gradient mutation point.

[0040] Among them, the signal-to-noise ratio (Signal-to-Noise Ratio, SNR) is an important indicator to measure the quality of the signal, indicating the ratio of signal strength to noise strength. Higher SNR means that the signal is more obvious relative to the noise, and the signal quality is better; lower SNR indicates that the noise has greater interference on the signal, and the signal is more difficult to identify. In the fields of communication, radar, audio, etc., SNR is a key indicator to evaluate the performance of the system, which is often used to measure the clarity and reliability of the signal.

[0041] In a possible implementation, S5 specifically includes sub-steps S501 and S502: S501: When the signal-to-noise ratio SNR satisfies , the bandwidth is expanded to a frequency domain range that is expanded by 20% outside the gradient mutation point.

[0042] It should be noted that in a low signal-to-noise ratio environment, the signal may be weak and easily overwhelmed by noise. By expanding the bandwidth, the key signal components of the signal can be better captured, avoiding signal loss due to too narrow bandwidth. At the same time, expanding the bandwidth helps to improve the robustness of filtering, so that more signal features can be retained and identified, thereby improving the signal recognition ability in a low signal-to-noise ratio situation.

[0043] S502: When the signal-to-noise ratio SNR satisfies , the bandwidth is contracted to a frequency domain range that is contracted by 10% inside the gradient mutation point.

[0044] It should be noted that in a high signal-to-noise ratio environment, the influence of noise is small, and the interference of irrelevant frequency components can be reduced by contracting the bandwidth. This helps to more accurately extract the key features of the signal, avoiding unnecessary noise and irrelevant frequency information from being included, thereby improving the accuracy and efficiency of signal processing, while also reducing the computational burden.

[0045] The prior art has no energy gradient analysis and directly relies on a fixed bandwidth. The present application locates the signal effective bandwidth boundary through the gradient mutation point, providing a basis for adaptive filtering.

[0046] In the embodiments of the present application, a bandwidth and signal-to-noise ratio negative feedback adjustment mechanism is adopted, which can dynamically adjust the filtering bandwidth according to the time-frequency characteristics of the signal. When the signal-to-noise ratio is high, the bandwidth can be contracted to avoid unnecessary noise interference, while when the signal-to-noise ratio is low, the bandwidth is automatically expanded to ensure that the key features of the signal are not lost, effectively improving the signal recognition accuracy under low signal-to-noise ratio, enhancing the robustness of signal processing, reducing signal loss due to bandwidth mismatch, and improving the reliability in low interception target detection. Ultimately, more accurate signal recognition and modulation pattern classification are achieved.

[0047] S6: Based on the filtering bandwidth boundary, the radar pulse signal is filtered in bandwidth.

[0048] Specifically, dynamic bandwidth filtering is achieved by an FIR filter, which preserves key features such as instantaneous frequency.

[0049] The prior art uses fixed bandwidth filtering, which is prone to recognition failure due to noise falling within the bandwidth or signal overflowing the bandwidth in a low signal-to-noise ratio environment. The present application realizes dynamic matching of the filtering bandwidth and the signal time-frequency characteristics through energy gradient extraction and negative feedback bandwidth adjustment. Test data shows that for low intercepted signals with a signal-to-noise ratio ≤-5dB, the signal-to-noise ratio is improved by ≥15dB, and the signal saturation distortion tolerance is improved by 300%, solving the problem that traditional fixed bandwidth cannot adapt to complex electromagnetic environments.

[0050] S7: Calculate the instantaneous frequency, first-order differential frequency and second-order differential frequency of the filtered signal through Hilbert transform, and form a feature vector.

[0051] Among them, Hilbert transform is a mathematical transform that converts a real-valued signal into its corresponding complex-valued signal, generating an "analysis signal" of the signal. Its core function is to construct a virtual part with a phase difference of 90 degrees from the original signal, thereby obtaining the instantaneous amplitude and instantaneous frequency of the signal.

[0052] S8: According to the feature vector, predict the clustering number range through a long short-term memory neural network.

[0053] Among them, the long short-term memory neural network (LSTM) is a special type of recurrent neural network (RNN) designed to solve the gradient vanishing or explosion problem of traditional RNN when processing long sequence data. LSTM introduces memory cells and control gate mechanisms to selectively save and forget information, thereby capturing long-term dependencies.

[0054] The clustering number of the prior art needs to be manually preset, and the present application dynamically predicts the clustering number range through LSTM, avoiding the deviation caused by manual setting. LSTM has the ability to process long sequence data and capture long-term dependencies, and can predict the clustering number based on the signal features at the previous moment, thereby effectively reducing the deviation and inaccuracy caused by manually setting the clustering number in traditional methods. This method improves the adaptability and accuracy of clustering analysis, especially in complex and variable signal environments, it can flexibly adjust the clustering strategy according to the characteristics of the actual signal, further improving the performance and robustness of signal recognition.

[0055] S9: According to the feature vector, perform clustering analysis within the clustering number range through the K-Means clustering algorithm.

[0056] Among them, K-Means clustering algorithm is a commonly used unsupervised learning method, which is used to divide the data set into K different clusters, and the center of each cluster is the mean value of the samples in the cluster. The algorithm iteratively selects K initial cluster centers at random, then assigns samples to the nearest cluster according to the distance between the sample and the cluster center, and then updates the center of each cluster, and repeats the process until the cluster center no longer changes.

[0057] In one possible implementation, S9 specifically includes: S901: Initialize different numbers of cluster centers within the range of cluster numbers.

[0058] S902: Cluster according to the cluster center by K-Means clustering algorithm.

[0059] S903: Compare the silhouette coefficients of the clustering analysis results under different cluster numbers.

[0060] Among them, the silhouette coefficient is an index to measure the quality of clustering, which combines the intra-cluster tightness and inter-cluster separation.

[0061] S904: Output the clustering analysis result with the maximum silhouette coefficient.

[0062] It should be noted that by initializing different numbers of cluster centers within the range of cluster numbers and using K-Means clustering algorithm for clustering analysis, and then comparing the silhouette coefficients of the clustering results under different cluster numbers, the most appropriate cluster number can be effectively selected. This method avoids the bias of manually setting the cluster number, and uses the silhouette coefficient as an index to measure the quality of clustering, ensuring that the selected cluster number can best reflect the true structure of the data. By outputting the clustering result with the maximum silhouette coefficient, the accuracy and stability of clustering analysis can be improved, thereby enhancing the precision and reliability of signal modulation pattern recognition.

[0063] The prior art relies on single-dimensional features or manually preset cluster numbers, resulting in high error rate under low signal-to-noise ratio. The present application uses three-dimensional feature extraction (instantaneous frequency + differential frequency) and LSTM clustering prediction, combined with parallel clustering optimization, to form a double guarantee in feature dimension and clustering rationality. Under the condition of signal-to-noise ratio-5dB, the modulation recognition accuracy is improved from 62% of the prior art to 89%, significantly improving the reliability of signal sorting and threat assessment.

[0064] S10: Output the modulation pattern label according to the clustering analysis result.

[0065] Optionally, the modulation pattern label includes: linear frequency modulation and phase encoding.

[0066] In a possible implementation, the signal intra-pulse modulation pattern identification method further includes: using a CPU and GPU parallel processing architecture to process the short-time Fourier transform, bandwidth filtering and clustering analysis in parallel.

[0067] The prior art is based on a CPU serial architecture, and the single-pulse processing delay is often more than hundreds of milliseconds, which cannot meet the needs of high-density signal scenarios. The application realizes efficient use of computing resources through GPU parallel optimization (CUDA Kernel splitting, pulsatile pipeline, shared memory).

[0068] Further, the pulsatile pipeline can be designed, and the STFT calculation of the next frame is started when clustering the previous frame.

[0069] Further, the preprocessed data (such as the time-frequency matrix) is stored in the global memory of the GPU, and is transmitted through the shared memory of the thread block, thereby reducing the CPU-GPU data copy delay.

[0070] It should be noted that the application simultaneously supports CPU and GPU deployment, improves signal detectability through adaptive filtering, improves recognition accuracy through multi-dimensional clustering, and improves real-time response capability through parallel architecture, thereby forming a full-link optimization from signal detection to feature extraction, and providing key technical support for electronic countermeasures and intelligence acquisition in complex battlefield environments.

[0071] Embodiment 2: Referring to Figure 5 The signal intra-pulse modulation pattern identification system 20 provided by the embodiment of the application includes a processor 201 and a memory 202.

[0072] The memory 202 stores programs or instructions executable on the processor 201, and the programs or instructions are executed by the processor 201 to implement the steps of the signal intra-pulse modulation pattern identification method of embodiment 1, and achieve the same technical effects.

[0073] In the description of the embodiments of the present application, it needs to be understood that the terms "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "center", "top", "bottom", "top", "bottom", "inner", "outer", "inner side", "outer side" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. Among them, "inside" refers to the inside or enclosed area or space. "Periphery" refers to the area around a particular component or a particular area.

[0074] In the description of the embodiments of the present application, the terms "first", "second", "third", "fourth" are only used for description purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second", "third", "fourth" can be explicitly or implicitly included one or more. In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0075] In the description of the embodiments of the present application, it needs to be explained that, unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connecting", "assembling" should be understood in a broad sense, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be directly connected, or indirectly connected through an intermediate medium; it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0076] In the description of the embodiments of the present application, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0077] In the description of the embodiments of the present application, it needs to be understood that "-" and "~" represent the range of the same of two numerical values, and the range includes the end points. For example: "A-B" represents a range greater than or equal to A and less than or equal to B. "A~B" represents a range greater than or equal to A and less than or equal to B.

[0078] In the description of the embodiments of the present application, the term "and / or" herein is only a description of the association relationship between the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " herein generally represents that the front and rear associated objects have an "or" relationship.

[0079] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A method for recognizing intra-pulse modulation patterns, characterized in that, include: S1: Receive radar pulse signals; S2: Perform a short-time Fourier transform on the radar pulse signal to generate a time-frequency matrix; S3: Calculate the energy distribution gradient at each time point along the frequency axis in the time-frequency matrix; S4: Identify gradient abrupt change points based on the energy distribution gradient at each time point; S5: A negative feedback adjustment mechanism for bandwidth and signal-to-noise ratio is adopted to determine the filter bandwidth boundary based on the gradient abrupt change point; S6: Based on the filtering bandwidth boundary, perform bandwidth filtering on the radar pulse signal; S7: Calculate the instantaneous frequency, first-order differential frequency, and second-order differential frequency of the filtered signal using Hilbert transform, and form a feature vector. S8: Based on the feature vector, predict the range of the number of clusters using a long short-term memory neural network; S9: Based on the feature vector, perform cluster analysis within the range of the number of clusters using the K-Means clustering algorithm; S10: Output modulation style labels based on cluster analysis results.

2. The signal pulse modulation pattern recognition method according to claim 1, characterized in that, The following is included after S1 and before S2: The radar pulse signal is converted from analog to digital.

3. The signal pulse modulation pattern recognition method according to claim 1, characterized in that, S2 specifically includes: S201: Dynamically adjust the window length of the sliding time window according to the signal carrier frequency; S202: Calculate the frequency components of each window using short-time Fourier transform to construct a three-dimensional time-frequency matrix containing time, frequency, and energy.

4. The signal pulse modulation pattern recognition method according to claim 3, characterized in that, S201 specifically includes: S2011: Before processing each frame of signal, perform spectral analysis on the previous frame signal using Fast Fourier Transform to extract the main peak frequency. S2012: Using the main peak frequency as the current carrier frequency value, compare it with the initial carrier frequency to calculate the carrier frequency deviation; S013: When the carrier frequency deviation satisfies At the same time, the window length of the sliding time window remains unchanged; S014: When the carrier frequency deviation satisfies When necessary, adjust the window length of the sliding time window; S015: When the carrier frequency deviation satisfies When this happens, the window length is reset, and the initial value setting is re-executed to ensure that the window length of the time window is adapted to the new carrier frequency.

5. The signal pulse modulation pattern recognition method according to claim 1, characterized in that, Specifically, S3 is: The energy distribution gradient is calculated by the energy difference between adjacent frequency points.

6. The signal pulse modulation pattern recognition method according to claim 1, characterized in that, S5 specifically includes: S501: When the signal-to-noise ratio (SNR) satisfies At that time, the bandwidth is extended to a frequency domain range 20% beyond the gradient abrupt change point; S502: When the signal-to-noise ratio (SNR) satisfies At that time, the bandwidth is reduced to a frequency domain range that is 10% smaller than the gradient abrupt change point.

7. The signal pulse modulation pattern recognition method according to claim 1, characterized in that, S9 specifically includes: S901: Within the range of the number of clusters, initialize different numbers of cluster centers; S902: Clustering is performed according to the cluster centers using the K-Means clustering algorithm; S903: Silhouette coefficients for comparing cluster analysis results with different numbers of clusters; S904: Output the cluster analysis result with the largest silhouette coefficient.

8. The signal pulse modulation pattern recognition method according to claim 1, characterized in that, The modulation style labels include: linear frequency modulation and phase coding.

9. The signal pulse modulation pattern recognition method according to claim 1, characterized in that, Also includes: It uses a CPU and GPU parallel processing architecture to perform short-time Fourier transform, bandwidth filtering, and cluster analysis in parallel.

10. A signal pulse modulation pattern recognition system, characterized in that, include: Processor and memory; The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the signal pulse modulation pattern recognition method as described in any one of claims 1 to 9.

Images