Sonar pulse interception method and product based on MCP architecture

Abstract

The invention belongs to the technical field of signal interception and passive sonar systems, and relates to a sonar pulse interception method and product based on an MCP architecture, and the method comprises the steps: obtaining a to-be-analyzed sonar pulse signal; based on an MCP architecture, a trained Mama model is adopted as a time sequence data processing core, long-time dependence features of a sonar pulse signal to be analyzed are extracted, and then the long-time dependence features are input into a large language model to obtain higher-level feature understanding; converting the generated feature understanding into a feature, fusing the feature with a long-time dependent feature, and obtaining a waveform category and a signal parameter of the sonar pulse signal to be analyzed, thereby realizing interception and classification of the sonar pulse signal. The method has high operability and good application prospects, and can be popularized and applied to incoming wave interception tasks in other fields.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal interception and passive sonar systems, and in particular relates to a sonar pulse interception method and product based on an MCP architecture. Background Art

[0002] To intercept sonar pulse signals, the task should include waveform classification, waveform parameter estimation, and interception result analysis. The parameters that need to be estimated include pulse frequency, bandwidth, pulse width, and repetition period. When the relative motion between the sonar platform and the receiving platform is unknown, the only parameters that can be estimated are the frequency, bandwidth, pulse width, and repetition period affected by Doppler. Traditionally, this approach involves a step-by-step approach using multiple signal processing methods. First, energy detection is used to identify the time period during which the incoming signal exists. Then, by converting the signal to the frequency domain, the frequency with the highest energy is determined based on the spectral peak. The instantaneous frequency is extracted using the Hilbert transform, or parameter estimation is performed using time-frequency analysis methods such as the short-time Fourier transform (STFT), fractional Fourier transform (FRFT), and Wigner-Ville distribution (WVD). This processing requires combining multiple signal processing techniques and setting multiple thresholds based on various rules for comprehensive judgment. The cascaded classification and parameter estimation tasks present the following challenges: 1. Error accumulation: Classification errors can also affect parameter estimation; and the sequential estimation of multiple parameters can also cause errors to propagate. 2. Computational redundancy: Classification and regression models run independently, resulting in high computational complexity. 3. Independent optimization of tasks leads to insufficient information utilization. Classification and regression tasks are actually coupled, and multiple parameters affect each other. Sequential processing will inevitably lead to information loss.

[0003] To address these issues, researchers have begun exploring smarter multi-task optimization strategies. Multi-Task Cooperative Processing (MCP) is a computational framework developed based on Multi-Task Learning (MTL). It emphasizes the collaborative learning and optimization of multiple related tasks, improving overall task performance by sharing information and features. In the fields of deep learning and signal processing, MCP is often used to simultaneously execute multiple tasks, such as classification, regression, detection, and segmentation, enabling them to mutually enhance each other and achieve better results. Summary of the Invention

[0004] The purpose of the present invention is to overcome the defects of the prior art and propose a sonar pulse interception method and product based on the MCP architecture.

[0005] In view of this, the present invention proposes a sonar pulse interception method based on the MCP architecture, comprising:

[0006] Obtaining sonar pulse signals to be analyzed;

[0007] Based on the MCP architecture, the trained Mamba model is used as the core of time series data processing to extract the long-term dependency features of the sonar pulse signal to be analyzed, and then input into the large language model to obtain a higher level of feature understanding;

[0008] The generated feature understanding is converted into features and then fused with long-term dependent features to obtain the waveform category and signal parameters of the sonar pulse signal to be analyzed, thereby realizing the interception and classification of the sonar pulse signal.

[0009] Preferably, the input of the Mamba model is the sonar pulse signal to be analyzed, and the output is the embedded representation.

[0010] Preferably, the input of the large language model is an embedding representation, and a text description is output through feature projection and normalization.

[0011] Preferably, the waveform categories include: CW, LFM, HFM and combined waveforms; the signal parameters include: frequency, bandwidth, pulse width and delay.

[0012] Preferably, the method further comprises a Mamba model training step, comprising:

[0013] Step 1: Construct a dataset of sonar pulse signals through simulation and generate labels to obtain a training set;

[0014] Step 2: Preprocess the data in the training set;

[0015] Step 3: Randomly shuffle the preprocessed data and divide it into several epochs. In each epoch, set the batch size based on the computing power. traverse the training data and input it into the Mamba model. Then, input the embedded representation output by the Mamba model into the large language model. Use the joint loss function to calculate the gradient layer by layer and update the parameters for supervised learning until a trained Mamba model is obtained.

[0016] Preferably, the step 1 comprises:

[0017] Step 1-1: For a transmitted signal s(t), the data received by a hydrophone is x(t). Build a signal model that includes the effects of multipath, reverberation, and noise:

[0018]

[0019] Among them, α i is the complex propagation coefficient, τ i is the delay of the i-th echo, K is the total number of echoes, ζ iis the Doppler coefficient, r(t) is reverberation, n(t) is white noise, and t represents the time domain;

[0020] Step 1-2: Randomly set the waveform type to obey the discrete uniform distribution;

[0021] Set the original frequency to the possible frequency range, and set the bandwidth, pulse width and period to obey discrete uniform distribution;

[0022] Set the radial velocity to obey uniform distribution in the possible range and obtain the corresponding Doppler coefficient;

[0023] The frequency, bandwidth, pulse width and period affected by Doppler are calculated based on the original frequency, bandwidth, pulse width, period and Doppler coefficient;

[0024] The number of multipaths and delay times are randomly generated and sorted, with the first delay being the direct wave and having a normalized amplitude. The attenuation factors of the other paths are set within a set range.

[0025] Randomly set the signal-to-noise ratio to 10, 20 or 30dB, and the signal-to-mixing ratio to 10, 20 or 30dB;

[0026] Steps 1-3: Randomly generate several samples and construct training sets, validation sets, and test sets according to a certain ratio;

[0027] Steps 1-4: Set up the storage structure, including: file name, waveform type, sampling rate, original starting frequency, bandwidth, pulse width, period, Doppler velocity, frequency affected by Doppler, bandwidth, pulse width, period, number of multipaths, and delay; if there is no pulse signal, the waveform type is set to 0, and all parameters except the sampling rate are set to empty, thereby generating a label.

[0028] Preferably, the step 2 comprises:

[0029] Step 2-1: Use a unified sampling frequency and standardize the data according to the maximum possible length. For data with insufficient length, add Gaussian noise at the end to achieve time domain normalization;

[0030] Step 2-2: According to the frequency range of the pulse signal, the effective frequency range is intercepted according to the set coefficient to achieve frequency domain normalization;

[0031] Step 2-3: Perform short-time Fourier transform to obtain the time-frequency spectrum of the signal. At the same time, normalize the time domain waveform and spectrum data, and compress the fast Fourier transform results and short-time Fourier transform results using a logarithmic scale to enhance the feature expression capability, thereby achieving normalization of the time-frequency domain.

[0032] Step 2-4: Based on the time-frequency features processed in step 2-3, further normalize the key parameters of the signal, including frequency, bandwidth, pulse width, and delay, unify the dimensions, and improve the consistency of the parameters, thereby achieving the standardization of the label parameters.

[0033] Preferably, the joint loss function in step 3 is:

[0034]

[0035] in, is the total loss function, and are the loss functions for classification, regression, and inference respectively, λ1, λ2, and λ3 are the corresponding weighting coefficients, and λ3 is a variable coefficient.

[0036] In another aspect, the present invention provides a computer program product, comprising a computer program, which implements the steps of the above method when executed by a processor.

[0037] Compared with the prior art, the advantages of the present invention are:

[0038] This paper innovatively proposes a sonar pulse signal interception method based on a multi-task collaborative processing framework, enabling pulse signal classification and parameter estimation. This method effectively extracts multi-layer waveform features using the Mamba model, resulting in higher efficiency, feature sharing, and reduced computational effort. This method also allows for deeper information mining, collaborative multi-task processing, and unified optimization of the entire task flow, reducing error accumulation and achieving end-to-end optimization. This approach is highly operational and has promising application prospects, and can be extended to other wave interception tasks in other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a method flow chart of a sonar pulse interception method based on the MCP architecture of the present invention;

[0040] Figure 2 shows the training process, where Figure 2(a) shows how the training loss and validation loss change with the number of training epochs, Figure 2(b) shows how the training accuracy and validation accuracy change with the number of training epochs, Figure 2(c) shows how the prediction accuracy of different parameters (such as fo_acc, B_acc, T_acc, and delay_acc) changes with the number of training epochs, and Figure 2(d) shows how the relative errors of different parameters (such as fo_rel, B_rel, T_rel, and delay_rel) change with the number of training epochs.

[0041] Figure 3 is the Confusion Matrix;

[0042] Figure 4 shows the parameter estimation, which shows the relationship between the true value and the predicted value when the model predicts different parameters. Figure 4(a) is the prediction result of the f0 parameter, Figure 4(b) is the prediction result of the B parameter, Figure 4(c) is the prediction result of the T parameter, and Figure 4(d) is the prediction result of the delay parameter. DETAILED DESCRIPTION

[0043] To better solve the problem of pulse signal interception, we proposed a multi-task framework that uses the same feature extraction network for classification, regression, and reasoning. Classification can help regression reduce the search scope and improve accuracy.

[0044] Aiming at intercepting sonar pulses, the present invention designs a classification and parameter estimation method for unknown sonar pulse signals based on a multi-task framework, including:

[0045] Obtaining sonar pulse signals to be analyzed;

[0046] Based on the MCP architecture, the trained Mamba model is used as the core of time series data processing to extract the long-term dependency features of the sonar pulse signal to be analyzed, and then input into the large language model to obtain a higher level of feature understanding;

[0047] The generated feature understanding is converted into features and then fused with long-term dependent features to obtain the waveform category and signal parameters of the sonar pulse signal to be analyzed, thereby realizing the interception and classification of the sonar pulse signal.

[0048] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0049] Example 1

[0050] Embodiment 1 of the present invention provides a sonar pulse interception method based on the MCP architecture, such as Figure 1 As shown, the steps are as follows:

[0051] MCP extracts the global timing information of the signal by adopting a structure of shared feature layers and task-specific branches, and then implements three branch tasks: classification branch (output waveform category), parameter regression branch (output signal parameters), and interception analysis branch (output whether the interception conditions are met).

[0052] 1. Dataset Construction

[0053] a) Constructing a dataset of sonar pulse signals through simulation

[0054] Assume that the data received by a hydrophone is x(t). Considering the environment in which the sonar pulse is received, a signal model including the effects of multipath, reverberation, and noise is established as follows:

[0055]

[0056] The received data is composed of several path signals superimposed with reverberation r(t) and white noise n(t), while the pulse signal echoes are all transmitted by a sonar platform through different paths. i is the complex propagation coefficient, its amplitude is the attenuation of the echo, and its phase includes the phase of the high-frequency carrier and the phase change of the signal during propagation. i is the time delay of the echo. Due to the radial velocity between the transmitting platform and the receiving platform, the pulse signals arriving from each path are affected by the Doppler effect, and there is a Doppler coefficient ζ i .

[0057] b) Waveform type

[0058] The system is configured based on commonly used sonar pulses, supporting CW, LFM, HFM, and combined waveforms, and generates different samples based on random parameters. The waveform parameters can be set based on possible bandwidth, pulse width, and period.

[0059] Considering the possible relative motion between the sound source and the receiving platform, the Doppler coefficient is set according to the relative velocity v Where c is the speed of sound, and the frequency is adjusted to match different motion conditions.

[0060] c) Multipath effect

[0061] Multipath includes direct waves, paths arriving via the sea surface, and paths arriving via the seabed. Multiple propagation paths with varying delays and amplitudes simulate the arrival of signals along these different paths. The direct wave should have the smallest delay and the largest amplitude. The delays of other paths can be set based on the simulated ocean environment, the depth of the sound source, and the receiving platform. The amplitude difference between other paths and the direct wave can be set using Lambert's empirical formula, among other factors, depending on factors such as the grazing angle, interface smoothness, and seabed texture.

[0062] d) Reverberation

[0063] In monitoring mode, after receiving a direct wave, reverberation persists for a long time due to the influence of various scatterers on the sea surface, seabed, and water column. This reverberation may persist due to the influence of the previous cycle. Reverberation is created by convolving random coherent noise and spectrum expansion with the transmitted pulse. Reverberation can also be constructed based on a spatial model with multiple scattering points.

[0064] e) Noise

[0065] Gaussian white noise is set to simulate the influence of ocean environmental noise and self-noise on the receiving platform.

[0066] f) Simulation parameters

[0067] The waveform type is randomly set to follow a discrete uniform distribution. The original frequency, bandwidth, pulse width, and period are set to follow a discrete uniform distribution within the constraints. The radial velocity is set to follow a uniform distribution within the possible range, and the corresponding Doppler coefficient is obtained. The frequency, bandwidth, pulse width, and period affected by Doppler are calculated based on the original frequency, bandwidth, pulse width, period, and Doppler coefficient. The number of multipaths and delay times are randomly generated and sorted, with the first delay being the direct wave and having a normalized amplitude. The attenuation factors of the other paths are set between 0.2 and 0.7 (adjustable based on actual conditions). The signal-to-noise ratio and signal-to-mixing ratio are randomly set to 10 / 20 / 30dB and 10 / 20 / 30dB, respectively.

[0068] g) Generate data

[0069] 10,000 samples are randomly generated and the training, validation, and test sets are constructed in a ratio of 8:1:1. Each sample is saved as a bin file.

[0070] h) Generate tags

[0071] The storage structure contains information such as the file name, waveform type, sampling rate, original start frequency, bandwidth, pulse width, period, Doppler velocity, frequency affected by Doppler, bandwidth, pulse width, period, number of multipath paths, and delay. It is stored as a .csv file.

[0072] For CW signals, the bandwidth is set to the reciprocal of the pulse width. Since the pulse width range is set from 20ms to 1s, the bandwidth range is 1Hz to 50Hz. For non-CW signals, the minimum bandwidth is 200Hz.

[0073] In the case where there is no pulse signal, the waveform type is set to 0, and all parameters except the sampling rate are set to empty.

[0074] 2. Data Preprocessing

[0075] a) Time domain normalization

[0076] The sampling frequency is unified and the data is sized to the maximum possible length. For data with insufficient length, very low amplitude Gaussian noise is added at the end.

[0077] b) Frequency domain normalization

[0078] Assume that the frequency range of the pulse signal is f1-f2. After performing FFT on the data, the effective frequency range is intercepted according to f1*0.5 to f2*1.2.

[0079] c) Time-frequency domain normalization

[0080] In order to take into account both frequency and time resolution, a higher overlap coefficient is set and short-time Fourier transform is performed to obtain the time-frequency spectrum.

[0081] Normalize time-domain waveform and spectrum data to facilitate processing by large models and deep learning models.

[0082] FFT and STFT results are processed using a logarithmic scale to enhance the dynamic range.

[0083] d) Label parameter normalization

[0084] The values of these four parameters vary widely, with frequencies ranging from thousands of Hz to tens of milliseconds, and time delays potentially ranging from microseconds to tens of milliseconds. The bandwidth also depends on the signal type: for CW signals, it's within tens of Hz (depending on the pulse width); for non-CW signals, it's greater than 200 Hz. Normalization methods are designed based on the statistical characteristics of each parameter, taking into account the varying distribution of bandwidth parameters across signal types.

[0085] 3. MCP Task Construction

[0086] For sonar pulse signals, the Mamba model is used as the core architecture for time series data processing to extract long-term dependency features. While interception result analysis should be prioritized, classification and parameter estimation should complement each other.

[0087] a) Define the MCP task flow

[0088] Create MCP tasks, including signal classification, parameter estimation, and inference.

[0089] b) Joint optimization of multi-branch tasks

[0090] Mamba is used as the core for time series data processing within a multi-task collaborative framework. Large models (such as DeepSeek) are introduced for decision enhancement: embeddings generated by Mamba are fed into the large model to obtain a higher level of feature understanding. Waveform patterns are analyzed through natural language inference mechanisms to generate auxiliary decisions. A joint loss function is used to simultaneously optimize classification, regression, and inference tasks:

[0091]

[0092] in, is the total loss function, and are the loss functions for classification, regression, and inference, respectively, while λ1, λ2, and λ3 are their weighting coefficients. λ3 is a variable coefficient that is set to 0 in the first few epochs of training.

[0093] 4. Mamba Processing

[0094] The Mamba model is used as the core architecture for time series data processing to extract long-term dependency features.

[0095] a) Mamba model feature extraction

[0096] Input: time domain waveform, FFT spectrum, STFT time-frequency spectrum.

[0097] Task: Extract features and generate embedding representations.

[0098] b) Training the Mamba model:

[0099] Supervised learning: Use simulation datasets to train waveform classification tasks (such as CW, LFM, HFM, and combination waveforms).

[0100] Parameter regression: Train the Mamba model to predict parameters such as frequency, bandwidth, arrival time, and pulse width.

[0101] The output header includes a classification header (predicted waveform type) and a parameter header (predicted signal parameters f0, B, T, delay)

[0102] 5. LLM (Large Language Model) interface processing

[0103] a) Feature extraction

[0104] Receive the features extracted by the Mamba model, and then perform feature projection and normalization.

[0105] b)LLM call

[0106] Build a prompt, call an external or local LLM API, and process the returned text description.

[0107] c) Feature Fusion

[0108] The descriptions generated by LLM are converted into features and fused with Mamba features.

[0109] 6. Training Process

[0110] a) Batch processing

[0111] Randomly shuffle the training data and perform 100 epochs of training. In each epoch, set the batch size based on the computing power and traverse the training data.

[0112] b) Forward propagation

[0113] After passing through the various components of the model, the final prediction output is generated.

[0114] c) Backpropagation

[0115] Backpropagation starts with loss calculation, then calculates gradients layer by layer and updates parameters.

[0116] 7. Testing and Evaluation

[0117] a) Performance evaluation

[0118] Comprehensively evaluate performance through confusion matrix analysis, accuracy of each category, misclassification analysis, and parameter prediction performance

[0119] b) Visual analysis

[0120] Visualize training history, including loss curves, accuracy curves, and parameter prediction errors; visualize prediction results, including waveform type predictions, parameter prediction scatter plots, and error distribution plots.

[0121] The results of classification and parameter estimation are given based on the test data.

[0122] Example 2

[0123] Example 2 of the present invention provides a sonar pulse interception method based on the MCP architecture, specifically comprising: constructing a simulation experiment: constructing a simulation data set according to the above method, and performing training, verification, and testing. The results are as follows.

[0124] The estimates for the two groups of samples in the test set are as follows:

[0125] Table 1

[0126] Signal Type frequency bandwidth Pulse Width Latency Actual value CW+LFM 4919.85 1297.85 0.812343 0.001946 Estimated value CW+LFM 4887.33 1289.69 0.831973 0.002267 0.66％ 0.63％ 2.42％ 16.48％

[0127] Table 2

[0128] Signal Type frequency bandwidth Pulse Width Latency Actual value CW 5663.12 2.16 0.462445 0.004700 Estimated value CW 5645.56 2.11 0.474499 0.004917 0.315％ 2.61％ 2.61％ 4.61％

[0129] Figure 2 shows the training process, where Figure 2(a) shows how the training loss and validation loss change with the epoch, which is used to evaluate the model's fitting and generalization capabilities; Figure 2(b) shows how the training accuracy and validation accuracy change with the epoch, reflecting the model's performance in classification tasks; Figure 2(c) shows how the prediction accuracy of different parameters (such as fo_acc, B_acc, T_acc, delay_acc) changes with the epoch, evaluating the model's performance in multi-target prediction tasks; Figure 2(d) shows how the relative error of different parameters (such as fo_rel, B_rel, T_rel, delay_rel) changes with the epoch, measuring the accuracy of the model's predictions.

[0130] like Figure 3 The following is a confusion matrix, where the horizontal axis represents the predicted value (Predicted) and the vertical axis represents the true value (True). The value in each cell represents the model's classification result for the corresponding category, with dark blue representing a higher number of samples. Specifically, the values on the diagonal (e.g., 2538, 2416, 2419, 2512) represent the number of samples correctly classified by the model, while the values on the other off-diagonal lines (e.g., 71, 43, 1) indicate instances where the model misclassified the sample.

[0131] Figure 4 shows the parameter estimation, which demonstrates the relationship between the true value and the predicted value when the model predicts different parameters. Figure 4(a) shows the prediction result of the f0 parameter. The scatter points in the figure represent the corresponding relationship between the true value and the predicted value, and the red dotted line is the ideal consistency line (that is, the predicted value is equal to the true value). Figure 4(b) shows the prediction result of the B parameter, which also demonstrates the accuracy of the prediction through scatter points and a consistency line. Figure 4(c) shows the prediction result of the T parameter, which further evaluates the model's prediction ability for this parameter. Figure 4(d) shows the prediction result of the delay parameter, which reflects the model's performance in the delay prediction task through scatter point distribution and a consistency line.

[0132] Example 3

[0133] The embodiment of the present invention may further provide a computer program product, including a computer program. When the computer program is executed by a processor, each step in the above method embodiment can be implemented.

[0134] Finally, it should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention and are intended to be encompassed by the claims of the present invention.

Claims

1. A sonar pulse interception method based on the MCP architecture, comprising: Obtaining sonar pulse signals to be analyzed; Based on the MCP architecture, the trained Mamba model is used as the core of time series data processing to extract the long-term dependency features of the sonar pulse signal to be analyzed, and then input into the large language model to obtain a higher level of feature understanding; The generated feature understanding is converted into features and then fused with long-term dependent features to obtain the waveform category and signal parameters of the sonar pulse signal to be analyzed, thereby realizing the interception and classification of the sonar pulse signal.

2. The sonar pulse interception method based on the MCP architecture according to claim 1 is characterized in that: The input of the Mamba model is the sonar pulse signal to be analyzed, and the output is the embedded representation.

3. The sonar pulse interception method based on the MCP architecture according to claim 1 is characterized in that: The input of the large language model is the embedding representation, and through feature projection and normalization, the output is a text description.

4. The sonar pulse interception method based on the MCP architecture according to claim 1 is characterized in that: The waveform categories include: CW, LFM, HFM and combined waveforms; the signal parameters include: frequency, bandwidth, pulse width and delay.

5. The sonar pulse interception method based on the MCP architecture according to claim 1 is characterized in that: The method also includes a Mamba model training step, including: Step 1: Construct a dataset of sonar pulse signals through simulation and generate labels to obtain a training set; Step 2: Preprocess the data in the training set; Step 3: Randomly shuffle the preprocessed data and divide it into several epochs. In each epoch, set the batch size based on the computing power. traverse the training data and input it into the Mamba model. Then, input the embedded representation output by the Mamba model into the large language model. Use the joint loss function to calculate the gradient layer by layer and update the parameters for supervised learning until a trained Mamba model is obtained.

6. The sonar pulse interception method based on the MCP architecture according to claim 5 is characterized in that: The step 1 comprises: Step 1-1: For a transmitted signal s(t), the data received by a hydrophone is x(t). Build a signal model that includes the effects of multipath, reverberation, and noise: Among them, α i is the complex propagation coefficient, τ i is the delay of the i-th echo, K is the total number of echoes, ζ i is the Doppler coefficient, r(t) is reverberation, n(t) is white noise, and t represents the time domain; Step 1-2: Randomly set the waveform type to obey the discrete uniform distribution; Set the original frequency to the possible frequency range, and set the bandwidth, pulse width and period to obey discrete uniform distribution; Set the radial velocity to obey uniform distribution in the possible range and obtain the corresponding Doppler coefficient; The frequency, bandwidth, pulse width and period affected by Doppler are calculated based on the original frequency, bandwidth, pulse width, period and Doppler coefficient; The number of multipaths and delay times are randomly generated and sorted, with the first delay being the direct wave and having a normalized amplitude. The attenuation factors of the other paths are set within a set range. Randomly set the signal-to-noise ratio to 10, 20 or 30dB, and the signal-to-mixing ratio to 10, 20 or 30dB; Steps 1-3: Randomly generate several samples and construct training sets, validation sets, and test sets according to a certain ratio; Steps 1-4: Set up the storage structure, including: file name, waveform type, sampling rate, original starting frequency, bandwidth, pulse width, period, Doppler velocity, frequency affected by Doppler, bandwidth, pulse width, period, number of multipaths, and delay; if there is no pulse signal, the waveform type is set to 0, and all parameters except the sampling rate are set to empty, thereby generating a label.

7. The sonar pulse interception method based on the MCP architecture according to claim 5 is characterized in that: The step 2 includes: Step 2-1: Use a unified sampling frequency and standardize the data according to the maximum possible length. For data with insufficient length, add Gaussian noise at the end to achieve time domain normalization; Step 2-2: According to the frequency range of the pulse signal, the effective frequency range is intercepted according to the set coefficient to achieve frequency domain normalization; Step 2-3: Perform short-time Fourier transform to obtain the time-frequency spectrum of the signal. At the same time, normalize the time domain waveform and spectrum data, and compress the fast Fourier transform results and short-time Fourier transform results using a logarithmic scale to enhance the feature expression capability, thereby achieving normalization of the time-frequency domain. Step 2-4: Based on the time-frequency features processed in step 2-3, further normalize the key parameters of the signal, including frequency, bandwidth, pulse width, and delay, unify the dimensions, and improve the consistency of the parameters, thereby achieving the standardization of the label parameters.

8. The sonar pulse interception method based on the MCP architecture according to claim 5 is characterized in that: The joint loss function in step 3 is: in, is the total loss function, and are the loss functions for classification, regression, and inference respectively, λ1, λ2, and λ3 are the corresponding weighting coefficients, and λ3 is a variable coefficient.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to claim 1 are implemented.

Images