A broadband signal intelligent detection and identification method and system

Abstract

The present application relates to modulation signal detection technical field, especially in a kind of wideband signal intelligent detection identification method and system, receive to be identified wideband intermediate frequency signal;Utilize short-time Fourier transform to be identified wideband signal is transformed to time-frequency domain, obtain to be identified wideband signal amplitude spectrum and phase spectrum;And utilize HSV color model to carry out fusion processing to signal amplitude spectrum and phase spectrum, obtain the time-frequency spectrogram of joint amplitude spectrum and phase spectrum, the HSV color model is constructed according to signal amplitude spectrum and phase spectrum;Based on time-frequency spectrogram detection and identify the burst signal in to be identified wideband signal, and the detection identification result of burst signal is visualized display.The present application can improve the accuracy rate of target signal under complex noise background being detected and identified, can be aimed at the burst signal in wideband data and carry out real-time intelligent detection and modulation mode identification, so that in variable electromagnetic environment intelligent analysis of signal becomes possible.

Description

Technical Field

[0001] The present invention relates to the technical field of modulation signal detection, and in particular to a method and system for intelligent detection and identification of broadband signals. Background Art

[0002] In the current field of signal processing, broadband data acquisition has become an essential means of acquiring large amounts of data, especially in scenarios requiring real-time monitoring and analysis of complex signal environments. Burst signals in broadband data often carry valuable information, making real-time detection and identification of these signals particularly crucial. Therefore, developing efficient and real-time signal detection and identification algorithms holds significant research and practical significance. Currently, signal modulation identification techniques fall into several main areas. The first is based on likelihood decision theory, including the average likelihood ratio test (ALRT), the generalized likelihood ratio test (GLRT), and the improved ALRT algorithm. ALRT calculates the likelihood function of the received signal and classifies the modulation mode based on these values. GLRT and the improved ALRT are suitable for scenarios with long data lengths or multi-antenna reception, reducing computational complexity by optimizing the likelihood ratio test. The second is the traditional feature extraction method: including classification based on the instantaneous characteristics of the signal, such as using the instantaneous amplitude, phase and frequency of the signal to identify the modulation mode; classification based on the signal constellation diagram (IQ data) features, by analyzing the IQ diagram features of the signal, extracting information related to the modulation mode; classification based on the high-order cumulants of the signal, using the characteristic that the high-order cumulants of the noise are zero, and identifying the modulation mode through the high-order cumulants of the signal; classification based on the cyclic spectrum characteristics of the signal, using the cyclic stationary characteristics of the signal, and identifying the modulation mode through the cyclic spectrum characteristics of the signal. The third is the method based on deep learning: including extracting the effective information features contained in the signal image based on the Convolutional Neural Network (CNN), and using these features to identify the modulation mode; extracting the signal time series features based on the Recurrent Neural Network (RNN) and identifying the modulation mode.

[0003] While traditional signal detection methods can meet certain requirements to a certain extent, they have limitations in terms of real-time performance and accuracy when processing large amounts of data and complex electromagnetic environments. These limitations are often affected by factors such as the channel environment and high computational complexity. In the field of signal modulation recognition, traditional modulation recognition methods based on likelihood decision theory and feature extraction, while performing well in certain scenarios, require a high level of prior knowledge. This leads to poor robustness and limited recognition accuracy when faced with complex signals and changing environments. Deep learning techniques, particularly neural network-based algorithms, address this issue to a certain extent by analyzing the characteristics of the signal set to be identified and automatically learning and extracting highly discriminative features. However, most current neural network-based algorithms primarily analyze the characteristics of the signal set to be identified, selecting a single, highly discriminative signal feature or directly combining features from multiple domains as network input, or optimizing the network structure to extract deeper features to improve performance. This often overlooks the complementary nature of features extracted from different modalities and different classifiers. Summary of the Invention

[0004] To this end, the present invention provides a method and system for intelligent detection and recognition of broadband signals, which solves the problem that the signal recognition accuracy is affected when facing complex signals and changing environments in existing signal detection and recognition.

[0005] According to the design scheme provided by the present invention, on the one hand, a method for intelligent detection and identification of broadband signals is provided, comprising:

[0006] receiving a broadband intermediate frequency signal to be identified;

[0007] The broadband signal to be identified is transformed into the time-frequency domain using short-time Fourier transform to obtain the amplitude spectrum and phase spectrum of the broadband signal to be identified; the amplitude spectrum and phase spectrum of the signal are fused using the HSV color model to obtain a time-frequency spectrum diagram of the combined amplitude spectrum and phase spectrum. The HSV color model is constructed based on the amplitude spectrum and phase spectrum of the signal;

[0008] Detect and identify burst signals in the broadband signal to be identified based on the time-frequency spectrum diagram, and visualize the detection and identification results of the burst signals.

[0009] As the broadband signal intelligent detection and identification method of the present invention, further, receiving a broadband intermediate frequency signal to be identified includes:

[0010] A broadband signal acquisition card is used to receive the broadband intermediate frequency signal to be identified and pre-process the broadband intermediate frequency signal. The broadband signal acquisition card acquires the broadband intermediate frequency signal to be identified based on an orthogonal two-channel IQ signal data acquisition and storage system.

[0011] As the broadband signal intelligent detection and recognition method of the present invention, further, the broadband signal to be identified is transformed into the time-frequency domain by using short-time Fourier transform, comprising:

[0012] Obtaining a complex time-frequency matrix of the broadband signal to be identified based on short-time Fourier transform, and obtaining an amplitude spectrum and a phase spectrum of the broadband signal to be identified based on the complex time-frequency matrix;

[0013] The phase spectrum is preprocessed with phase jump using the triangular transformation method, and a phase spectrum with symmetrical texture features is obtained.

[0014] As the broadband signal intelligent detection and recognition method of the present invention, the signal amplitude spectrum and phase spectrum are further fused using the HSV color model, including:

[0015] The signal amplitude is used as saturation, the sine value of the phase spectrum trigonometric function is used as chromaticity, and the cosine value of the phase spectrum trigonometric function is used as brightness to obtain the HSV color model;

[0016] The HSV color model is used to fuse the amplitude spectrum and phase spectrum to obtain a time-frequency spectrum of the joint phase spectrum and amplitude spectrum.

[0017] As the broadband signal intelligent detection and identification method of the present invention, further, based on the time-frequency spectrum diagram, the burst signal in the broadband signal to be identified is detected and identified, including:

[0018] The signal detection problem is transformed into a regression problem by using a pre-trained single-stage detector to detect and identify burst signals in the broadband signal to be identified. The single-stage detector is constructed using a single-stage target detection algorithm.

[0019] As the broadband signal intelligent detection and recognition method of the present invention, further, the single-stage detector adopts a lightweight YOLOv5 network structure, and the lightweight YOLOv5 network structure includes a backbone network for extracting multi-scale feature data of the input image based on a feature pyramid, a neck network for fusing the multi-scale feature data, and a head network for performing target detection and recognition based on the multi-scale fused features.

[0020] As the broadband signal intelligent detection and recognition method of the present invention, further, the single-stage detector training process includes:

[0021] Using signal-to-noise ratio parameters to simulate wireless communication environment signals and generate corresponding broadband time-spectrum graph sample data, and label the sample data with label information, the label including modulation mode, signal center time, signal duration, signal carrier frequency and signal bandwidth;

[0022] At each signal-to-noise ratio, the sample data is divided into training set, validation set and test set according to the preset ratio;

[0023] The training set is used to train the one-stage detector, and the validation set and test set are used to adjust parameters and evaluate the performance of the trained one-stage detector.

[0024] On the other hand, the present invention also provides a broadband signal intelligent detection and recognition system, comprising: a signal receiving module, a feature fusion module and a target detection module, wherein:

[0025] A signal receiving module, used for receiving a broadband intermediate frequency signal to be identified;

[0026] A feature fusion module is used to transform the broadband signal to be identified into the time-frequency domain using short-time Fourier transform to obtain the amplitude spectrum and phase spectrum of the broadband signal to be identified; and to fuse the amplitude spectrum and phase spectrum of the signal using the HSV color model to obtain a time-frequency spectrum diagram of the combined amplitude spectrum and phase spectrum. The HSV color model is constructed based on the amplitude spectrum and phase spectrum of the signal;

[0027] The target detection module is used to detect and identify burst signals in the broadband signal to be identified based on the time-frequency spectrum diagram, and to visualize the detection and identification results of the burst signals.

[0028] Beneficial effects of the present invention:

[0029] The present invention adopts an HSV-based amplitude-phase feature fusion method to make the time-frequency spectrum features of the signal have a richer latent space representation, significantly improve the feature contrast between the signal and noise, and improve the accuracy of detection and identification of the target signal in a complex noise background, and can perform real-time intelligent detection and modulation mode identification for burst signals in broadband data; the target detection network is applied to the field of signal detection and identification, showing obvious advantages in the real-time and accuracy of detection and identification. Experimental data verify that when the broadband signal-to-noise ratio is higher than -10dB, the accuracy and recall rate are both maintained at above 90%, and it has better detection and identification effect for weak signals. It can realize end-to-end processing capability, real-time response speed, automatic feature learning capability, low power consumption and computing requirements in one-stop signal intelligent real-time analysis, reduce the complexity and cost of hardware deployment, and make it possible to perform intelligent analysis of signals in a changeable electromagnetic environment. It has good application prospects in the field of signal intelligent detection and classification and identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a schematic diagram of the broadband signal intelligent detection and identification process in the embodiment;

[0031] Figure 2 Schematic diagram of the QPSK time-frequency amplitude spectrum (a) and phase spectrum (b) in the embodiment;

[0032] Figure 3 This is a schematic diagram of the YOLOv5s model network structure in the embodiment;

[0033] Figure 4 Schematic diagram of the accuracy (a) and recall (b) curves of the model under various signal-to-noise ratio data in the embodiment. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and technical solutions.

[0035] In view of the problems that traditional signal detection has limitations in terms of real-time performance and accuracy when processing large amounts of data and complex electromagnetic environments, and is easily affected by factors such as channel environment and high computational complexity, the embodiment of the present invention provides a method for intelligent detection and identification of broadband signals, such as Figure 1 As shown, including:

[0036] S101: Receive a broadband intermediate frequency signal to be identified.

[0037] Specifically, a broadband signal acquisition card is used to receive and pre-process the broadband intermediate frequency signal to be identified. The broadband signal acquisition card acquires the broadband intermediate frequency signal to be identified based on an orthogonal two-channel IQ signal data acquisition and storage system. The orthogonal two-channel IQ signal data acquisition and storage system can be implemented using an existing inter-channel data acquisition system.

[0038] S102. Use short-time Fourier transform to transform the broadband signal to be identified into the time-frequency domain to obtain the amplitude spectrum and phase spectrum of the broadband signal to be identified; and use the HSV color model to fuse the signal amplitude spectrum and phase spectrum to obtain a time-frequency spectrum diagram of the combined amplitude spectrum and phase spectrum, wherein the HSV color model is constructed based on the signal amplitude spectrum and phase spectrum.

[0039] The method of transforming the broadband signal to be identified into the time-frequency domain using short-time Fourier transform may include:

[0040] Obtaining a complex time-frequency matrix of the broadband signal to be identified based on short-time Fourier transform, and obtaining an amplitude spectrum and a phase spectrum of the broadband signal to be identified based on the complex time-frequency matrix;

[0041] The phase spectrum is preprocessed with phase jump using the triangular transformation method, and a phase spectrum with symmetrical texture features is obtained.

[0042] For amplitude and phase modulation signals, the time-frequency characteristics of the signal are mainly affected by the phase and amplitude, and the time-frequency spectrogram shows more levels and changes. The time-frequency spectrogram of signals with different amplitude and phase modulation methods does not have obvious difference characteristics, so it is not possible to accurately identify the modulation method of each amplitude and phase signal only by relying on the time-frequency amplitude spectrum characteristics. In order to more comprehensively describe the time-frequency characteristics of the signal, in the embodiments of the case, the phase characteristics are added on the basis of the original amplitude spectrum characteristics, and the joint phase spectrum and amplitude spectrum time-frequency spectrogram is used for signal modulation method identification to improve the identification accuracy.

[0043] The time-domain signal S is subjected to STFT to obtain a complex time-frequency matrix Where T and F correspond to the number of time frames and frequency points. The amplitude spectrum and phase spectrum corresponding to the time-frequency matrix can be represented as:

[0044]

[0045] P = ∠S ∈ [-π, π] T×F

[0046] Where ∠ is the radian operation, such as Figure 2 As shown in the amplitude spectrum and phase spectrum diagram of the QPSK signal.

[0047] Since the time-frequency matrix S will compress the continuous phase of (-∞, +∞) into a phase matrix within the range of [-π, +π] during the phase calculation, the phase will jump between +π and -π, which will cause the phase jump and phase ambiguity on the phase spectrogram, and the phase spectrum will lack the symmetry texture features compared to the amplitude spectrum. This is the main reason why the amplitude spectrum in the time-frequency spectrogram is selected as the signal feature in most cases.

[0048] In order to reduce the influence of phase jump and phase ambiguity on the phase spectrum, in the embodiments of the case, the triangular transformation method is used to obtain the phase spectrum with symmetric texture features, and the phase values of +π and -π are the same after triangular transformation, which overcomes the influence of phase jump. The phase spectrum after triangular transformation can be represented as:

[0049] P cos = cos(P) ∈ [-1, 1] T×F

[0050] P sin = sin(P) ∈ [-1, 1] T×F

[0051] Where cos, sin are the cosine and sine transformation.

[0052] In the field of signal processing, especially in environments with low signal-to-noise ratios, the impact of noise on phase becomes particularly significant, which directly affects the time-frequency spectrum's ability to detect and identify signal targets. In this embodiment, the signal amplitude spectrum and phase spectrum are fused using the HSV color model, which can be designed to include:

[0053] The signal amplitude is used as saturation, the sine value of the phase spectrum trigonometric function is used as chromaticity, and the cosine value of the phase spectrum trigonometric function is used as brightness to obtain the HSV color model;

[0054] The HSV color model is used to fuse the amplitude spectrum and phase spectrum to obtain a time-frequency spectrum of the joint phase spectrum and amplitude spectrum.

[0055] By introducing the HSV color model, the amplitude spectrum and phase spectrum are effectively fused, aiming to enhance the characteristic contrast between the signal and the noise. The HSV color model, namely the hue, saturation, and value model, processes the time-frequency spectrum through color, saturation, and value. In the embodiment of this case, the amplitude of the signal is regarded as saturation (S), the sine value of the trigonometric function is used to represent the hue (H), and the cosine value corresponds to the value (V). Through mapping, the amplitude not only determines the strength of the signal, but also controls the saturation of the color in the time-frequency spectrum. Specifically, by adjusting the saturation, the significance of the signal phase in the entire time-frequency spectrum can be enhanced, while effectively reducing the interference of noise on the signal detection and recognition process.

[0056] By fusing the HSV color model, the characteristic contrast between the signal and the noise can be significantly improved, making the signal target easier to detect and identify in a complex noise background; because the importance of the signal phase is enhanced, the signal target can be identified more accurately, even when the signal-to-noise ratio is low, thus improving the detection accuracy; by controlling the saturation, the influence of noise on signal detection and recognition is reduced, the robustness of signal processing is improved, and the influence of noise is reduced, providing new tools and ideas for the field of signal processing.

[0057] S103 : Detect and identify a burst signal in the to-be-identified broadband signal based on the time-frequency spectrum diagram, and visualize the detection and identification results of the burst signal.

[0058] Specifically, a pre-trained single-stage detector can be used to transform the signal detection problem into a regression problem to detect and identify burst signals in the broadband signal to be identified. The single-stage detector is constructed using a single-stage target detection algorithm.

[0059] Single-stage object detection algorithms have gained significant traction in the field of object detection due to their superior computational efficiency and simplified model architecture. They can quickly predict the bounding box and category of an object directly from an image, eliminating the need for candidate region generation. This significantly reduces computational effort and increases speed, making them suitable for applications requiring high real-time performance. Furthermore, the end-to-end training of single-stage detectors simplifies the training process and enhances model consistency. Their excellent generalization and adaptability enable them to maintain stable performance across diverse datasets and changing environments. Furthermore, single-stage detectors are easy to integrate and scale, reducing hardware costs and energy consumption and increasing the feasibility of large-scale deployments.

[0060] Among them, the single-stage detector can adopt a lightweight YOLOv5 network structure, which includes a backbone network for extracting multi-scale feature data of the input image based on a feature pyramid, a neck network for fusing the multi-scale feature data, and a head network for target detection and recognition based on the multi-scale fused features.

[0061] In the current field of signal processing, broadband data acquisition has become an important means of acquiring large amounts of data, especially when real-time monitoring and analysis of complex signal environments is required. Burst signals in broadband data often carry important information, making real-time detection and identification of these signals particularly important. Therefore, exploring efficient and real-time signal detection and identification algorithms has significant research value and practical significance.

[0062] In this embodiment, the single-stage target detection algorithm YOLOv5 is selected to achieve rapid recognition of objects in the image. An end-to-end regression problem is used to solve the target detection task. The position and category information of all objects in the image can be output through a single forward pass, which can improve the detection speed and meet the application scenarios with real-time requirements.

[0063] like Figure 3 As shown in the figure, this lightweight network structure mainly includes a backbone network, a feature pyramid network, a neck network, and a head network. Due to its low parameter count and computational complexity, the model can complete object detection tasks in a short time, which is very important for real-time application scenarios or resource-constrained devices. At the same time, the YOLOv5 model is highly convenient in use and deployment, allowing for the rapid construction of customized object detection models and their migration and application in different application scenarios.

[0064] The single-stage detector training process can be designed to include:

[0065] The signal-to-noise ratio parameter is used to simulate the wireless communication environment signal and generate corresponding wideband time-frequency spectrum sample data, and the sample data is labeled with label information, including modulation mode, signal center time, signal duration, signal carrier frequency and signal bandwidth.

[0066] The sample data is divided into a training set, a validation set and a test set according to a preset proportion at each signal-to-noise ratio.

[0067] The training set is used to train the single-stage detector, and the validation set and the test set are used to adjust parameters and evaluate performance of the trained single-stage detector.

[0068] The sample signal set is {2FSK, 4FSK, 8FSK, BPSK, QPSK, 8PSK, 16QAM, 32QAM, 64QAM, 128QAM, 16APSK, MSK, 4CPFSK}. By setting multiple signal-to-noise ratio parameters, the diversity and authenticity of signals in a wireless communication environment are simulated. At each signal-to-noise ratio, the training set, the validation set and the test set respectively contain {420, 120, 60} wideband time-frequency spectrum graphs and {modulation mode, signal center time x, signal duration w, carrier frequency y, bandwidth h} labeled information to guide the YOLOv5 model to learn the features of signals on the time-frequency amplitude spectrum graph.

[0069] The wideband time-frequency spectrum graph parameters are shown in Table 1:

[0070] Table 1 Wideband time-frequency spectrum graph parameter setting table

[0071]

[0072] After the data set is prepared, the YOLOv5s model is trained using the data set, and the training parameters are shown in Table 2. After the model training is completed, the signal detection model is tested and comprehensively evaluated.

[0073] Table 2 Training parameter setting

[0074]

[0075] The performance of the model on the test set is shown in Table 2. According to the test results, the performance of the target detection model gradually improves with the increase of the signal-to-noise ratio and tends to converge to 1. In the case of high signal-to-noise ratio, accurate and comprehensive signal detection and recognition tasks can be performed based on the fusion time-frequency spectrum graph. When the signal-to-noise ratio is higher than -10dB, the accuracy and recall rate are both maintained above 90%. Figure 4

[0076] Further, based on the above method, the embodiment of the present application also provides a wideband signal intelligent detection and recognition system, comprising a signal receiving module, a feature fusion module and a target detection module, wherein,​

[0077] A signal receiving module, used for receiving a broadband intermediate frequency signal to be identified;

[0078] A feature fusion module is used to transform the broadband signal to be identified into the time-frequency domain using short-time Fourier transform to obtain the amplitude spectrum and phase spectrum of the broadband signal to be identified; and to fuse the amplitude spectrum and phase spectrum of the signal using the HSV color model to obtain a time-frequency spectrum diagram of the combined amplitude spectrum and phase spectrum. The HSV color model is constructed based on the amplitude spectrum and phase spectrum of the signal;

[0079] The target detection module is used to detect and identify burst signals in the broadband signal to be identified based on the time-frequency spectrum diagram, and to visualize the detection and identification results of the burst signals.

[0080] Unless otherwise specifically stated, the relative steps, numerical expressions and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present invention.

[0081] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0082] The units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person of ordinary skill in the art may use different methods to implement the described functions for each specific application, but such implementation is not considered to be beyond the scope of the present invention.

[0083] Those skilled in the art will appreciate that all or part of the steps in the above method can be performed by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a magnetic disk, or an optical disk. Alternatively, all or part of the steps in the above embodiment can be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or software functional modules. The present invention is not limited to any specific combination of hardware and software.

[0084] Finally, it should be noted that the above-described embodiments are merely specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, but not to limit the same. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that any person skilled in the art can still modify or easily think of changes to the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some of the technical features, within the technical scope disclosed by the present application. The modifications, changes or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for intelligent detection and identification of broadband signals, characterized in that: Include: receiving a broadband intermediate frequency signal to be identified; The broadband signal to be identified is transformed into the time-frequency domain using short-time Fourier transform to obtain the amplitude spectrum and phase spectrum of the broadband signal to be identified; the signal amplitude is used as saturation, the sine value of the phase spectrum trigonometric function is used as chroma, and the cosine value of the phase spectrum trigonometric function is used as brightness to obtain the HSV color model, and the amplitude spectrum and phase spectrum are fused using the HSV color model to obtain a time-frequency spectrum diagram of the combined amplitude spectrum and phase spectrum. The HSV color model is constructed based on the signal amplitude spectrum and phase spectrum; Based on the time-spectrogram and using a pre-trained single-stage detector, the signal detection problem is converted into a regression problem to detect and identify burst signals in the broadband signal to be identified, and the detection and identification results of the burst signals are visualized. The single-stage detector is constructed using a single-stage target detection algorithm.

2. The method for intelligent detection and identification of broadband signals according to claim 1, wherein: Receive the broadband intermediate frequency signal to be identified, including: A broadband signal acquisition card is used to receive the broadband intermediate frequency signal to be identified and pre-process the broadband intermediate frequency signal. The broadband signal acquisition card acquires the broadband intermediate frequency signal to be identified based on an orthogonal two-channel IQ signal data acquisition and storage system.

3. The method for intelligent detection and identification of broadband signals according to claim 1, wherein: The short-time Fourier transform is used to transform the broadband signal to be identified into the time-frequency domain, including: A complex time-frequency matrix of the broadband signal to be identified is obtained based on short-time Fourier transform, and an amplitude spectrum and a phase spectrum of the broadband signal to be identified are obtained based on the complex time-frequency matrix.

4. The method for intelligent detection and identification of broadband signals according to claim 1, wherein: The single-stage detector adopts a lightweight YOLOv5 network structure, which includes a backbone network for extracting multi-scale feature data of an input image based on a feature pyramid, a neck network for fusing the multi-scale feature data, and a head network for detecting and identifying targets based on the multi-scale fused features.

5. The method for intelligent detection and identification of broadband signals according to claim 1 or 4, characterized in that: The single-stage detector training process includes: Using signal-to-noise ratio parameters to simulate wireless communication environment signals and generate corresponding broadband time-spectrum graph sample data, and label the sample data with label information, the label including modulation mode, signal center time, signal duration, signal carrier frequency and signal bandwidth; At each signal-to-noise ratio, the sample data is divided into training set, validation set and test set according to the preset ratio; The training set is used to train the one-stage detector, and the validation set and test set are used to adjust parameters and evaluate the performance of the trained one-stage detector.

6. A broadband signal intelligent detection and recognition system, characterized in that: It includes: signal receiving module, feature fusion module and target detection module, among which, A signal receiving module, used for receiving a broadband intermediate frequency signal to be identified; A feature fusion module is used to transform the broadband signal to be identified into the time-frequency domain using a short-time Fourier transform to obtain the amplitude spectrum and phase spectrum of the broadband signal to be identified; and the signal amplitude is used as the saturation, the sine value of the phase spectrum trigonometric function as the chroma, and the cosine value of the phase spectrum trigonometric function as the brightness to obtain an HSV color model. The amplitude spectrum and the phase spectrum are fused using the HSV color model to obtain a time-frequency spectrum diagram of the combined amplitude spectrum and phase spectrum. The HSV color model is constructed based on the signal amplitude spectrum and phase spectrum; The target detection module is used to transform the signal detection problem into a regression problem based on the time-spectrogram and using a pre-trained single-stage detector to detect and identify burst signals in the broadband signal to be identified, and to visualize the detection and identification results of the burst signals. The single-stage detector is constructed using a single-stage target detection algorithm.

7. An electronic device, characterized in that: include: at least one processor, and a memory coupled to the at least one processor; The memory stores a computer program, and the computer program can be executed by the at least one processor to implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 5 can be implemented.

Images