Practical sampling short wave protocol sample data processing method and application

Abstract

The invention discloses a real-time acquisition short-wave protocol sample data processing method. The method comprises the following steps: acquiring a real-time acquisition short-wave protocol sample data set; sequentially intercepting burst signals in the actually-acquired short-wave protocol sample data set through an anti-fading energy detection method to form a burst signal set; if it is judged that the burst signal is a real number sample, the center frequency of the real number sample is obtained, and the real number sample is converted into an I / Q burst signal through down-conversion according to the center frequency; acquiring a sampling rate of the I / Q burst signal, selecting a corresponding low-pass filter according to the sampling rate, and filtering out-of-band redundant information of the I / Q burst signal through the low-pass filter; and calculating signal amplitudes and signal phases between the I path and the Q path of the I / Q burst signal in sequence, and forming data to be trained according to an amplitude and phase staggered arrangement mode. The problems of non-standard data specification, unbalanced data volume and high training difficulty of the actually-acquired short-wave communication protocol in the prior art can be solved.

Description

Technical Field

[0001] The present application relates to the field of communication and deep learning technology, and more specifically, to a method for processing actual shortwave protocol sample data and a shortwave protocol intelligent recognition model training device. Background Art

[0002] In the training and implementation of the intelligent shortwave protocol recognition model, the standardization of the actual protocol samples and the extraction of burst signals have always been the difficulties in the training of the recognition model. Specifically, due to the different sources and uses of the actual protocol samples, the starting position of the signal in the sample library is not fixed, the center frequency is not uniform, and the data specifications are different; the communication frequency of different shortwave protocols is different, resulting in an imbalance in the amount of data between different types of actual protocol samples; the multipath fading caused by the multipath effect of the actual protocol samples seriously affects the signal energy.

[0003] At present, among the commonly used processing methods for extracting and detecting burst signals from real-world protocol samples, manual screening and labeling is time-consuming and laborious; the synchronous demodulation method has a certain false detection rate and cannot extract protocol signals with unknown synchronization information. In view of the above situation, real-world protocol samples are difficult to be used as training data for recognition models in the actual training process. Therefore, it is an urgent problem to propose a standardized data processing method for real-world protocol samples and establish a suitable model training system based on the standardized data. Summary of the invention

[0004] In view of at least one defect or improvement need in the prior art, the present invention provides a method for processing actual shortwave protocol sample data and a shortwave protocol intelligent recognition model training device, aiming to solve the problems in the prior art of non-standard specifications of actual shortwave communication protocol data, unbalanced data volume and high training difficulty.

[0005] To achieve the above-mentioned purpose, according to the first aspect of the present invention, a method for processing real-sampled shortwave protocol sample data is provided, comprising: obtaining a real-sampled shortwave protocol sample data set; intercepting burst signals in the real-sampled shortwave protocol sample data set in sequence through an anti-fading energy detection method to form a burst signal set; if the burst signal in the burst signal set is determined to be a real number sample, obtaining the center frequency of the real number sample, and converting the real number sample into an I / Q burst signal through down-conversion according to the center frequency; obtaining the sampling rate of the I / Q burst signal, selecting a corresponding low-pass filter according to the sampling rate, and filtering out-of-band redundant information from the I / Q burst signal through the low-pass filter; and calculating the signal amplitude and signal phase between the I path and the Q path of the I / Q burst signal in sequence, and forming data to be trained in a staggered arrangement of amplitude and phase.

[0006] In one embodiment of the present invention, the acquisition of the actual shortwave protocol sample data set includes: a series of .wav format data received by a shortwave receiver in a wideband and sampled in a narrowband; the sample sampling rate adopts one of the standard gears of 9.6KHz, 24KHz, and 48KHz; the sample data indicates the signal bandwidth, and if the storage format is a real number signal, the signal center frequency is also indicated.

[0007] In one embodiment of the present invention, the anti-fading energy detection method includes the steps of: reading the actual shortwave protocol sample data to be detected and adding windows; calculating the signal energy under each window in turn and screening out a continuous relatively high-energy window sequence; for a relatively low-energy window sequence, loading a neural network that has been trained for anti-fading and performing energy compensation on it; re-screening out a continuous relatively high-energy window sequence to obtain a burst signal detected by energy.

[0008] In one embodiment of the present invention, the down-conversion formula is: MSG′(n)=MSG(n)×e -2πj×fc / fs×n ; Among them, MSG(n) is the real signal, MSG′(n) is the down-converted signal, fc is the signal center frequency, and fs is the sampling rate.

[0009] In one embodiment of the present invention, the relationship between the cutoff frequency of the low-pass filter and the sample sampling rate is: a sampling rate of 9.6KHz corresponds to a cutoff frequency of 3KHz; or a sampling rate of 24KHz corresponds to a cutoff frequency of 12KHz; or a sampling rate of 48KHz corresponds to a cutoff frequency of 24KHz.

[0010] In one embodiment of the present invention, the signal amplitudes of the I path and the Q path of the I / Q burst signal are expressed as: The signal phase is expressed as: S iA =arctan(S iQ / S iI ), where S iM is the signal amplitude, S iA is the signal phase, S iI is the I signal, S iQ It is the Q channel signal.

[0011] In one embodiment of the present invention, the arrangement of the signal amplitude and the signal phase is: 1M ,S 1A ,S 2M ,S 2A ,……,S nM ,S nA ; Where n is the number of shortwave protocol samples actually collected, S 1M ~S nM is the signal amplitude, S 1A ~S nAis the signal phase.

[0012] According to the second aspect of the present invention, there is also provided a shortwave protocol intelligent recognition model training device, comprising: a data loading module, used for batch reading actual shortwave protocol sample data under a specified directory, and performing data enhancement on protocol categories whose sample data is less than a preset number; a training parameter configuration module, used for adjusting the training hyperparameters and training modes required for the training process; a training module, used for initializing or reading a network recognition model, and adjusting the recognition model parameters through iterative training; a result output module, which generates a recognition accuracy change curve graph for monitoring the training process, and outputs a recognition result model.

[0013] In one embodiment of the present invention, the data enhancement includes: frequency domain enhancement: adding a random frequency deviation in the range of 200Hz to -200Hz to the sample data; time domain enhancement: adding blank data with a random length in the range of 0 to 75ms before and after the sample data; noise enhancement: superimposing random Gaussian noise with an intensity in the range of 0 to 2dB on the sample data.

[0014] In one embodiment of the present invention, the configuration of the training hyperparameters and training mode includes: the number of training rounds, which is used to adjust the number of iterative updates of model parameters; the number of training batches, which is used to configure the number of samples processed simultaneously in each round of training iteration; the learning rate base value, which is used to configure the degree of influence of each round of training iteration on parameter changes; the pre-trained model loading mode, which is used to configure whether to load the pre-trained model; the result saving mode, which is used to configure the way of saving the result model to save the best model or save all models.

[0015] On the other hand, an embodiment of the present invention further proposes an electronic device, comprising: a memory and one or more processors connected to the memory, the memory storing a computer program, and the processor being used to execute the computer program to implement the actual shortwave protocol sample data processing method as described in any one of the above embodiments.

[0016] On the other hand, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to execute the actual shortwave protocol sample data processing method as described in any one of the above embodiments.

[0017] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0018] The actual shortwave protocol sample data processing method and model training device provided in the embodiment of the present invention improve the shortwave protocol data by energy detecting the burst signal position, enhancing the protocol data of a small amount, etc., and performs specific training through a parameter-configurable training device, thereby improving the usability of the actual sample data in the shortwave protocol recognition model training task, thereby solving the current problems of non-standard specifications, unbalanced data volume and high training difficulty of actual shortwave communication protocol data. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A flowchart of a method for processing shortwave protocol sample data provided in an embodiment of the present application;

[0021] Figure 2 A structural diagram of a model training device provided in an embodiment of the present application;

[0022] Figure 3 This is a diagram of the parameter configuration interface of the model training device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0024] The terms "first", "second", "third", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0025] like Figure 1As shown, the first embodiment of the present invention proposes a method for processing real-sampled shortwave protocol sample data, including: step S1, obtaining a real-sampled shortwave protocol sample data set; step S2, intercepting the burst signals in the real-sampled shortwave protocol sample data set in turn through an anti-fading energy detection method to form a burst signal set; step S3, if it is determined that the burst signal in the burst signal set is a real number sample, then obtaining the center frequency of the real number sample, and converting the real number sample into an I / Q burst signal through down-conversion according to the center frequency; step S4, obtaining the sampling rate of the I / Q burst signal, selecting a corresponding low-pass filter according to the sampling rate, and filtering the I / Q burst signal through the low-pass filter to remove the out-of-band redundant information; step S5, sequentially calculating the signal amplitude and signal phase between the I path and the Q path of the I / Q burst signal, and forming the training data according to the staggered arrangement of the amplitude and the phase.

[0026] In step S1, for example, a shortwave receiver is used to obtain a real-time shortwave protocol sample data set in a designated controlled frequency band. Specifically, for example, a series of .wav format data is received in a wideband and sampled in a narrowband by a shortwave receiver; the sample sampling rate adopts one of the standard gears of 9.6KHz, 24KHz, and 48KHz; the sample data indicates the signal bandwidth, and if the storage format is a real number signal, the signal center frequency is also indicated.

[0027] In step S2, the acquired actual shortwave protocol sample data is used to detect the burst signal in the actual sample data using an anti-fading energy detection method, and the anti-fading energy detection method includes: reading the actual shortwave protocol sample data to be detected and adding windows; calculating the signal energy under each window in turn and screening out a continuous relatively high-energy window sequence; for a relatively low-energy window sequence, loading a neural network that has been trained for anti-fading and performing energy compensation on it; re-screening out a continuous relatively high-energy window sequence to obtain a burst signal detected by energy.

[0028] In step S3, if the actual burst signal is a real signal, it is necessary to read the center frequency of the signal and down-convert it to a baseband I / Q signal, as shown in the following formula:

[0029] MSG′(n)=MSG(n)×e -2πj×fc / fs×n

[0030] Among them, MSG(n) is a real signal, MSG′(n) is a down-converted signal, fc is the signal center frequency, and fs is the sampling rate.

[0031] In step S4, the sampling rate of the I / Q burst signal is obtained, and a corresponding low-pass filter is selected according to the sampling rate, and the I / Q burst signal is passed through the low-pass filter to filter out the out-of-band redundant information. The cutoff frequency and coefficient of the filter are determined by the sample sampling rate.

[0032] Optionally, the relationship between the cutoff frequency of the low-pass filter and the sample sampling rate is: a sampling rate of 9.6KHz corresponds to a cutoff frequency of 3KHz; or a sampling rate of 24KHz corresponds to a cutoff frequency of 12KHz; or a sampling rate of 48KHz corresponds to a cutoff frequency of 24KHz.

[0033] In step S5, the signal amplitude and signal phase between the I / Q burst signal I and Q are calculated in sequence, and the training data is formed in the arrangement of [amplitude-phase-amplitude-phase...]. The signal amplitude of the I / Q burst signal I and Q is expressed as: The signal phase is expressed as: s iA =arctan(S iQ / S iI ), where S iM is the signal amplitude, S iA is the signal phase, S iI is the I signal, S iQ It is the Q channel signal.

[0034] The arrangement of the signal amplitude and the signal phase is:

[0035] S 1M ,S 1A ,S 2M ,S 2A ,……,S nM ,S nA ;

[0036] Among them, n is the number of shortwave protocol samples actually collected, S 1M ~S nM is the signal amplitude, S 1A ~S nA is the signal phase.

[0037] In this way, the actual shortwave protocol sample data processing method provided by the embodiment of the present invention improves the shortwave protocol data by energy detecting the burst signal position, enhancing the protocol data of a small amount, etc., and performs specific training through a parameter-configurable training device, thereby improving the usability of the actual sample in the shortwave protocol recognition model training task, thereby solving the current problems of non-standard specifications, unbalanced data volume and high training difficulty of actual shortwave communication protocol data.

[0038] In addition, if Figure 2As shown, the second embodiment of the present invention proposes a shortwave protocol intelligent recognition model training device, for example, including: a data loading module, a training parameter configuration module, a training module and a result output module. Among them, the data loading module is used to batch read the actual shortwave protocol sample data under the specified directory, and perform data enhancement on the protocol category with less sample data than the preset number; the training parameter configuration module is used to adjust the training hyperparameters and training mode required for the training process; the training module is used to initialize or read the network recognition model, and adjust the recognition model parameters through iterative training; the result output module generates a recognition accuracy change curve for monitoring the training process, and outputs the recognition result model.

[0039] Specifically, the model training device is deployed on a high-performance server with a GPU, and its operation process includes the following steps:

[0040] Step 1: When the device is running, the user must first configure the training parameters on the user interface of the training device and select the training mode. The user's configuration information will be transmitted to the training module through the training parameter configuration module.

[0041] Step 2: The device will then batch read the training data under the specified configuration "training data directory" and process the training data through the data loading and processing module to form the data to be trained.

[0042] At this time, if the sample data is small (the amount of sample data of a certain type is less than 1000), the data enhancement module will be automatically started. The data enhancement module includes frequency domain enhancement, time domain enhancement and noise enhancement.

[0043] Frequency domain enhancement: add random frequency deviation in the range of 200Hz to -200Hz to the sample data;

[0044] The specific method is as follows:

[0045] MSG′(n)=MSG(n)×e -2πj(fc-offset) / fs×n

[0046] Wherein MSG(n) is the sample data to be enhanced, MSG′(n) is the enhanced sample data, fc is the signal center frequency, offset is the added random frequency deviation within the range of 200 Hz to -200 Hz, and fs is the sampling rate.

[0047] Time domain enhancement: add blank data with a random length of 0 to 75ms before and after the sample data;

[0048] Noise enhancement: random Gaussian noise with a strength range of 0 to 2dB is superimposed on the sample data.

[0049] Step 3: After the training data is processed, the device will automatically load the built-in model and initialize the model's internal network parameters. Then, it will automatically start training and iterating the model according to the configured training parameters until the maximum number of training rounds is reached. The progress of the training will be displayed in real time on the interactive interface.

[0050] Step 4: The result model generated by the training will be saved to the specified result output directory through the result saving module. The result model includes a training checkpoint model of the ".chkpt" type and an application model of the ".onnx" type. The training change curve and recognition accuracy confusion matrix generated by the training will also be saved to the result output directory through the result saving module.

[0051] Figure 3 This is a diagram of the parameter configuration interface of the model training device provided by the present invention. The model training device can interact with the user through the parameter configuration interface to control the training mode. The configurable parameters are as follows:

[0052] Training data directory: configure the storage location of the actual shortwave protocol data set. The directory location can be changed by writing in the window, or by clicking the "Select Directory" button to locate it.

[0053] Result save directory: Configure the storage location of the training results. The directory location can be changed by writing in the window, or by clicking the "Select Directory" button to locate it.

[0054] Number of training rounds: You can configure the number of training rounds to adjust the number of iterative updates of model parameters. The larger the value, the more times the training is performed.

[0055] Training batch number: You can configure the number of samples that are processed simultaneously in each training iteration. The larger the value, the higher the training efficiency and the better the convergence stability, but the more GPU memory is occupied.

[0056] Learning rate base value: You can configure the degree of influence of each training iteration on parameter changes. The larger the value, the faster the training speed, but the training is less likely to converge; the lower the value, the easier it is to converge, but the slower the training speed and the more likely it is to fall into the local optimum.

[0057] Pre-trained model loading mode: You can configure whether to load a pre-trained model. If it is "Yes", it means that the pre-trained model is loaded in this training, which is generally used when breakpoint resumption is required; if it is "No", it means training a new model.

[0058] Result saving mode: You can configure the result model saving method to save the best model or save all models. When it is "Best Model", it means that only the training result model with the best protocol classification accuracy is saved in each round of training; when it is "All Models", it means that all models are saved in each round of training.

[0059] The shortwave protocol intelligent recognition model training device disclosed in the second embodiment of the present invention is suitable for implementing the actual shortwave protocol sample data processing method as described in the first embodiment. The specific method can refer to the content of the first embodiment, so it will not be described in detail here. Optionally, the various modules and the above-mentioned other operations or functions in the second embodiment are respectively for implementing the method described in the first embodiment, and the beneficial effects of this embodiment are the same as those of the first embodiment, and for the sake of brevity, they are not repeated here.

[0060] The third embodiment of the present invention further proposes an electronic device, for example, comprising: at least one processing unit, and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit executes the method as described in the first embodiment, and the beneficial effects of the electronic device provided by this embodiment are the same as the beneficial effects of the actual shortwave protocol sample data processing method provided by the first embodiment.

[0061] The fourth embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when executed by a processor, and the beneficial effects of the computer-readable storage medium provided by this embodiment are the same as the beneficial effects of the actual shortwave protocol sample data processing method provided by the first embodiment.

[0062] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0063] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0064] In the several embodiments provided in the present application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0065] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0066] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0067] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, disk or optical disk and other media that can store program code.

[0068] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by entering a program to instruct related hardware. The program can be stored in a computer-readable memory, and the memory can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0069] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure here, those skilled in the art will easily think of the implementation scheme of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field not recorded in the present disclosure. The description and examples are regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

[0070] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0071] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for processing shortwave protocol sample data, characterized in that: The following steps are involved: Obtain the actual shortwave protocol sample data set; The burst signals in the actual shortwave protocol sample data set are intercepted in sequence by using an anti-fading energy detection method to form a burst signal set; If it is determined that the burst signal in the burst signal set is a real number sample, the center frequency of the real number sample is obtained, and the real number sample is converted into an I / Q burst signal through down-conversion according to the center frequency; Obtain the sampling rate of the I / Q burst signal, select a corresponding low-pass filter according to the sampling rate, and filter the I / Q burst signal through the low-pass filter to remove excess out-of-band information; The signal amplitude and signal phase between the I path and the Q path of the I / Q burst signal are calculated in sequence, and the training data is formed in a staggered arrangement of the amplitude and the phase.

2. The method for processing shortwave protocol sample data as claimed in claim 1, characterized in that: The acquisition of the actual shortwave protocol sample data set includes: A series of .wav format data received by a shortwave receiver in broadband and sampled in narrowband; The sample sampling rate adopts one of the standard gears of 9.6KHz, 24KHz, and 48KHz; The sample data indicates the signal bandwidth. If the saved format is a real signal, the signal center frequency is also indicated.

3. The method for processing shortwave protocol sample data as claimed in claim 1, characterized in that: The anti-fading energy detection method comprises the steps of: Read the actual shortwave protocol sample data to be detected and add window; Calculate the signal energy under each window in turn and filter out a sequence of continuous relatively high energy windows; For relatively low energy window sequences, load the neural network trained for anti-fading to perform energy compensation; The continuous relatively high energy window sequence is re-screened to obtain the burst signal detected by energy.

4. The method for processing shortwave protocol sample data as claimed in claim 1, characterized in that: The down-conversion formula is: MSG′(n)=MSG(n)×e -2πj×fc / fs×n ; Among them, MSG(n) is the real signal, MSG′(n) is the down-converted signal, fc is the signal center frequency, and fs is the sampling rate.

5. The method for processing shortwave protocol sample data as claimed in claim 1, characterized in that: The relationship between the cutoff frequency of the low-pass filter and the sample sampling rate is: The sampling rate is 9.6KHz, corresponding to the cutoff frequency is 3KHz; or A sampling rate of 24KHz corresponds to a cutoff frequency of 12KHz; or The sampling rate of 48KHz corresponds to a cutoff frequency of 24KHz.

6. The method for processing shortwave protocol sample data as claimed in claim 1, characterized in that: The signal amplitudes of the I path and the Q path of the I / Q burst signal are expressed as: The signal phase is expressed as: S iA =arctan(S iQ / S iI ), where S iM is the signal amplitude, S iA is the signal phase, S iI is the I signal, S iQ It is the Q channel signal.

7. The method for processing shortwave protocol sample data as claimed in claim 1, characterized in that: The arrangement of the signal amplitude and the signal phase is: S 1M ,S 1A ,S 2M ,S 2A ,……,S nM ,S nA ; Among them, n is the number of shortwave protocol samples actually collected, S 1M ~S nM is the signal amplitude, S 1A ~S nA is the signal phase.

8. A shortwave protocol intelligent recognition model training device, applicable to the shortwave protocol sample data processing method as claimed in any one of claims 1 to 7, characterized in that: include: The data loading module is used to batch read the actual shortwave protocol sample data in the specified directory and perform data enhancement on the protocol categories whose sample data is less than the preset number; The training parameter configuration module is used to adjust the training hyperparameters and training modes required for the training process; The training module is used to initialize or read the network recognition model and adjust the recognition model parameters through iterative training; The result output module generates a recognition accuracy change curve for monitoring the training process and outputs the recognition result model.

9. The shortwave protocol intelligent recognition model training device according to claim 8, characterized in that: The data enhancement includes: Frequency domain enhancement: add random frequency deviation in the range of 200Hz to -200Hz to the sample data; Time domain enhancement: add blank data with a random length of 0 to 75ms before and after the sample data; Noise enhancement: Random Gaussian noise with an intensity ranging from 0 to 2dB is superimposed on the sample data.

10. The shortwave protocol intelligent recognition model training device according to claim 8, characterized in that: The configuration of the training hyperparameters and training mode includes: The number of training rounds is used to adjust the number of iterative updates of model parameters; The number of training batches is used to configure the number of samples processed simultaneously in each training iteration; The basic value of the learning rate is used to configure the degree of influence of each training iteration on parameter changes; Pre-trained model loading mode, used to configure whether to load the pre-trained model; Result saving mode is used to configure the way to save the result model, either to save the best model or to save all models.

Images