Pulse neural network communication system identification method and system based on time-frequency feature fusion, electronic equipment and storage medium

Abstract

The invention discloses a pulse neural network communication system identification method and system based on time-frequency feature fusion, electronic equipment and a storage medium, and the method comprises the steps: obtaining communication system data from Internet of Things equipment, and making the communication system data into an initial communication system data set, dividing the initial communication system data set into training set data, verification set data and test set data according to a proportion; importing the divided training set into a time-frequency domain feature fusion module to obtain sample features of comprehensive time-frequency domain information; inputting the sample features of the comprehensive time-frequency domain information into a pulse neural network model to obtain an initial model predicted value and a model weight; substituting the model prediction value and the label value into a classification loss function for calculation to obtain a classification loss value; updating parameters of the spiking neural network model by using the classification loss value until an optimal spiking neural network model is obtained; according to the method, the communication system data detection efficiency and recognition precision in a complex real environment are improved, and meanwhile, the recognition energy consumption problem in a resource-limited scene is reduced.

Description

Technical Field

[0001] The present invention belongs to the field of communication security technology, and specifically relates to a pulse neural network communication format recognition method, system, electronic device and storage medium based on time-frequency feature fusion. Background Art

[0002] With the rapid growth of the Internet of Things and the widespread popularity of smart devices, it is expected that there will be more than 120 billion wireless devices by 2030. This surge in the number of devices has led to a sharp increase in demand for limited wireless spectrum resources, increasing the congestion of the spectrum and potentially causing signal interference, thereby affecting the quality and reliability of data transmission. Communication format recognition, as a key technology in the field of artificial intelligence, is used to detect, classify and manage different wireless communication signals, which is crucial in the face of an increasingly crowded wireless spectrum environment. Summary of the invention

[0003] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides a method for identifying communication formats using a pulse neural network based on time-frequency feature fusion, which can reduce the recognition energy consumption problem in resource-constrained scenarios while maintaining the recognition accuracy of the pulse neural network model with time-frequency feature fusion for communication format data.

[0004] Technical solution: In the first aspect, the present invention provides a pulse neural network communication mode recognition method based on time-frequency feature fusion, comprising:

[0005] Step 1: Obtain communication standard data from IoT devices and make an initial communication standard data set, and divide the initial communication standard data set into training set, verification set and test set data according to the proportion;

[0006] Step 2: Import the divided training set into the time-frequency domain feature fusion module to obtain the sample features of the comprehensive time-frequency domain information;

[0007] Step 3: Input the sample features of the comprehensive time-frequency domain information into the spiking neural network model to obtain the initial model prediction value and model weight;

[0008] Step 4: Substitute the model prediction value and label value into the classification loss function to obtain the classification loss value;

[0009] Step 5: Use the classification loss value to update the pulse neural network model parameters until the optimal pulse neural network model is obtained.

[0010] Furthermore, the specific steps of step 1 include:

[0011] Step 1.1: Use the communication standard collector to collect communication standard data from real IoT devices;

[0012] Step 1.2: Normalize the collected communication standard data into IQ sequence data and make it into a communication standard data set;

[0013] Step 1.3: Classify the IQ sequence data according to the preset label value, determine the label value of each IQ sequence data, and classify and store the IQ sequence data according to the label value to form an initial communication standard data set;

[0014] Step 1.4: Then divide the initial communication standard dataset into training set, test set and validation set in the ratio of 0.5:0.3:0.2.

[0015] Furthermore, the specific steps of step 2 include:

[0016] Step 2.1: Use Fourier transform on the divided training set to obtain FFT data with frequency domain characteristics;

[0017] Step 2.2: Import the initial communication format data and FFT format data into the time-frequency domain feature fusion module to obtain the sample features of the comprehensive time-frequency domain information.

[0018] Furthermore, the specific steps of step 3 include:

[0019] Step 3.1: Use the pulse neural network model built by the pulse neural network to fuse the time-frequency domain features and obtain the data features that comprehensively consider the time-frequency domain information;

[0020] Step 3.2: Input the data features into the classifier to obtain the model prediction value and the model weight value associated with the model prediction value.

[0021] Furthermore, the specific steps of step 4 include:

[0022] The calculation expression of the classification loss function is:

[0023]

[0024] In the formula, H(y i ,p i ) is the cross entropy loss function, L CE is the cross entropy loss function value, N is the total number of input categories, y i is the classification label value of the i-th sample, p i The model prediction value for the i-th sample.

[0025] Furthermore, the specific steps of step 5 include:

[0026] Step 5.1: Back propagate the classification loss value of the pulse neural network model fused with time-frequency features to obtain the gradient;

[0027] Step 5.2: Update the parameters of the pulse neural network model with time-frequency feature fusion according to the gradient until the optimal pulse neural network model with time-frequency feature fusion is obtained.

[0028] In a second aspect, the present invention provides a pulse neural network communication mode recognition system based on time-frequency feature fusion, comprising:

[0029] Data acquisition and processing module, update module, application module;

[0030] The data acquisition and processing module is used to obtain communication standard data and make it into an initial communication standard data set;

[0031] The updating module is used to import the training set obtained by dividing the initial communication standard data set into the pulse neural network model with time-frequency feature fusion to obtain the model prediction value and model weight; substitute the model prediction value and the label value into the classification loss function to calculate and obtain the classification loss value; and update the parameters of the pulse neural network model with time-frequency feature fusion according to the classification loss value until the optimal pulse neural network model with time-frequency feature fusion is obtained;

[0032] The application module is used to detect communication standard data online according to the pulse neural network model fused with time-frequency features, and adaptively identify communication standard data;

[0033] Among them, the update module is preset with training parameters of the pulse neural network model with time-frequency feature fusion, and the training parameters include: number of training times, batch size, optimizer parameters, learning rate, LIF attenuation factor, LIF reset mechanism and LIF pulse emission threshold.

[0034] In addition, the present invention also provides a computer storage medium and an electronic device, wherein a computer program is stored on the computer storage medium, and when the computer program is executed by a processor, the steps of the method of the present invention are implemented. The electronic device includes a memory and one or more processors, and the memory is used to store one or more programs; when the one or more programs are executed by one or more processors, the method of the present invention is implemented.

[0035] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0036] This method makes comprehensive use of the features of the time domain and frequency domain to obtain more comprehensive input sample information, thereby enhancing the robustness of communication format recognition. Secondly, the model utilizes the attention mechanism. Due to the characteristics of the attention mechanism, the sample features obtained are more discriminative and compress sample data with less information. At the same time, due to the characteristics of the pulse neural network, it can reduce power consumption and solve the problem of limited energy consumption in low-resource scenarios. Finally, this method is highly applicable and can adapt to different environments and types of communication format data, and has strong versatility and adaptability.

[0037] In various scenarios of complex real-world Internet data environments, this method creatively introduces signal processing, time-frequency fusion, and deep learning, solving the problems of low detection accuracy and high energy consumption of communication format data in recognition systems in complex scenarios with limited resources, thereby reducing the cost and difficulty of using the software. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flow chart of the pulse neural network communication system recognition method based on time-frequency feature fusion of the present invention;

[0039] Figure 2 It is a system model diagram of the pulse neural network model of time-frequency feature fusion of the present invention. DETAILED DESCRIPTION

[0040] In order to more fully understand the technical content of the present invention, the technical solution of the present invention is further introduced and illustrated in conjunction with specific embodiments below, but is not limited thereto.

[0041] Combination Figure 1 The pulse neural network communication mode recognition method based on time-frequency feature fusion of the present invention is further described, comprising:

[0042] Step 1: Obtain communication standard data from IoT devices and make an initial communication standard data set, and divide the initial communication standard data set into training set, verification set and test set data according to the proportion;

[0043] Step 2: Import the divided training set into the time-frequency domain feature fusion module to obtain the sample features of the comprehensive time-frequency domain information;

[0044] Step 3: Input the sample features of the comprehensive time-frequency domain information into the spiking neural network model to obtain the initial model prediction value and model weight;

[0045] Step 4: Substitute the model prediction value and label value into the classification loss function to obtain the classification loss value;

[0046] Step 5: Use the classification loss value to update the pulse neural network model parameters until the optimal pulse neural network model is obtained.

[0047] In this embodiment, the specific steps of step 1 include:

[0048] Step 1.1: Use the communication standard collector to collect communication standard data from real IoT devices;

[0049] Step 1.2: Normalize the collected communication standard data into IQ sequence data and make it into a communication standard data set;

[0050] Step 1.3: Classify the IQ sequence data according to the preset label value, determine the label value of each IQ sequence data, and classify and store the IQ sequence data according to the label value to form an initial communication standard data set;

[0051] Step 1.4: Then divide the initial communication standard dataset into training set, test set and validation set in the ratio of 0.5:0.3:0.2.

[0052] In this embodiment, the specific steps of step 2 include:

[0053] Step 2.1: Use Fourier transform on the divided training set to obtain FFT data with frequency domain characteristics;

[0054] Step 2.2: Import the initial communication format data and FFT format data into the time-frequency domain feature fusion module to obtain the sample features of the comprehensive time-frequency domain information.

[0055] In this embodiment, the specific steps of step 3 include:

[0056] Step 3.1: Use the pulse neural network model built by the pulse neural network to fuse the time-frequency domain features and obtain the data features that comprehensively consider the time-frequency domain information;

[0057] Step 3.2: Input the data features into the classifier to obtain the model prediction value and the model weight value associated with the model prediction value.

[0058] In this embodiment, the specific steps of step 4 include:

[0059] The calculation expression of the classification loss function is:

[0060]

[0061] In the formula, H(y i ,p i ) is the cross entropy loss function, L CE is the cross entropy loss function value, N is the total number of input categories, y i is the classification label value of the i-th sample, p i The model prediction value of the i-th sample; this embodiment updates the model through the prediction value and the training set label paper.

[0062] In this embodiment, the specific steps of step 5 include:

[0063] Step 5.1: Back propagate the classification loss value of the pulse neural network model fused with time-frequency features to obtain the gradient;

[0064] Step 5.2: Update the parameters of the pulse neural network model with time-frequency feature fusion according to the gradient until the optimal pulse neural network model with time-frequency feature fusion is obtained.

[0065] This embodiment further illustrates a pulse neural network communication mode recognition system based on time-frequency feature fusion in combination with experiments, including:

[0066] Data acquisition and processing module, update module, application module;

[0067] The data acquisition and processing module is used to obtain communication standard data and make it into an initial communication standard data set;

[0068] The updating module is used to import the training set obtained by dividing the initial communication standard data set into the pulse neural network model with time-frequency feature fusion to obtain the model prediction value and model weight; substitute the model prediction value and the label value into the classification loss function to calculate and obtain the classification loss value; and update the parameters of the pulse neural network model with time-frequency feature fusion according to the classification loss value until the optimal pulse neural network model with time-frequency feature fusion is obtained;

[0069] The application module is used to detect communication standard data online according to the pulse neural network model fused with time-frequency features, and adaptively identify communication standard data;

[0070] The updating module is pre-set with the training parameters of the pulse neural network model with time-frequency feature fusion, and the training parameters include: number of trainings, batch size, optimizer parameters, learning rate, LIF attenuation factor, LIF reset mechanism and LIF pulse emission threshold;

[0071] The initialization steps of the update module include:

[0072] A. Use the pulse neural network and pytorch code to build a pulse neural network model with time-frequency feature fusion. The loss function used in this model is the cross entropy loss function, and its formula is

[0073] B. Set the number of categories and training parameters required for training the basic model, input the training set data of the selected number of categories into the model for training, and obtain the optimal basic model; wherein the training parameters of the update module in this embodiment are preset as shown in the following table:

[0074] parameter value Training equipment GeForce GTX 3090Ti Training Environment Pytorch Optimizer Adam Batch size 64 Number of training sessions 200 LIF pulse transmission threshold 1.0 LIF Attenuation Factor 0.5 LIF reset mechanism Soft Reset

[0075] Table 1

[0076] This embodiment uses Fourier transform in communication format data containing 6 category labels to obtain FFT data with frequency domain information, and imports the initial communication format data and FFT data into the time-frequency sample level feature fusion module to obtain data features that integrate the time-frequency domain sample information. In the process of training the model, the best time-frequency fused pulse neural network model is obtained.

[0077] To summarize, the present invention maintains the recognition capability of the pulse neural network model based on time-frequency fusion for communication format data, and reduces the recognition energy consumption problem in resource-constrained scenarios while maintaining the recognition accuracy of the pulse neural network model based on time-frequency feature fusion for communication format data.

[0078] In various scenarios of complex real-world Internet data environments, signal processing, time-frequency fusion, and deep learning are creatively introduced to comprehensively improve the recognition system's ability to detect low accuracy and high energy consumption of communication format data in complex scenarios with limited resources, thereby reducing the cost and difficulty of using the software. This method is highly applicable and can adapt to different environments and types of communication format data, and has strong versatility and adaptability.

[0079] The embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0080] The embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0081] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0082] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0083] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0084] The above is only a preferred embodiment of the present invention. Without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A pulse neural network communication mode recognition method based on time-frequency feature fusion, characterized in that: include: Step 1: Obtain communication standard data from IoT devices and make an initial communication standard data set, and divide the initial communication standard data set into training set, verification set and test set data according to the proportion; Step 2: Import the divided training set into the time-frequency domain feature fusion module to obtain the sample features of the comprehensive time-frequency domain information; Step 3: Input the sample features of the comprehensive time-frequency domain information into the spiking neural network model to obtain the initial model prediction value and model weight; Step 4: Substitute the model prediction value and label value into the classification loss function to obtain the classification loss value; Step 5: Use the classification loss value to update the pulse neural network model parameters until the optimal pulse neural network model is obtained.

2. The pulse neural network communication system recognition method based on time-frequency feature fusion according to claim 1 is characterized in that: Step 1 specifically includes: Step 1.1: Use the communication standard collector to collect communication standard data from real IoT devices; Step 1.2: Normalize the collected communication standard data into IQ sequence data and make it into a communication standard data set; Step 1.3: Classify the IQ sequence data according to the preset label value, determine the label value of each IQ sequence data, and classify and store the IQ sequence data according to the label value to form an initial communication standard data set; Step 1.4: Then divide the initial communication standard dataset into training set, test set and validation set in the ratio of 0.5:0.3:0.

2.

3. The pulse neural network communication mode recognition method based on time-frequency feature fusion according to claim 1 is characterized in that: Step 2 specifically includes: Step 2.1: Use Fourier transform on the divided training set to obtain FFT data with frequency domain characteristics; Step 2.2: Import the initial communication format data and FFT format data into the time-frequency domain feature fusion module to obtain the sample features of the comprehensive time-frequency domain information.

4. The pulse neural network communication system recognition method based on time-frequency feature fusion according to claim 1 is characterized in that: Step 3 specifically includes: Step 3.1: Use the pulse neural network model built by the pulse neural network to fuse the time-frequency domain features and obtain the data features that comprehensively consider the time-frequency domain information; Step 3.2: Input the data features into the classifier to obtain the model prediction value and the model weight value associated with the model prediction value.

5. The pulse neural network communication system recognition method based on time-frequency feature fusion according to claim 2 is characterized in that: Step 4 specifically includes: Step 4.1: Substitute the model prediction value and label value into the classification loss function to obtain the classification loss value; the calculation expression of the classification loss function is: In the formula, H(y i ,p i ) is the cross entropy loss function, L CE is the cross entropy loss function value, N is the total number of input samples, y i is the classification label value of the i-th sample, p i The model prediction value for the i-th sample.

6. The pulse neural network communication system recognition method based on time-frequency feature fusion according to claim 1 is characterized in that: Step 5 specifically includes: Step 5.1: Back propagate the classification loss value of the pulse neural network model fused with time-frequency features to obtain the gradient; Step 5.2: Update the parameters of the pulse neural network model with time-frequency feature fusion according to the gradient until the optimal pulse neural network model with time-frequency feature fusion is obtained.

7. A pulse neural network communication mode recognition system based on time-frequency feature fusion, characterized in that: include: Data acquisition and processing module, update module, application module; The data acquisition and processing module is used to obtain communication standard data and make it into an initial communication standard data set; The updating module is used to import the training set obtained by dividing the initial communication standard data set into the pulse neural network model with time-frequency feature fusion to obtain the model prediction value and model weight; substitute the model prediction value and the label value into the classification loss function calculation to obtain the classification loss value; and updating the parameters of the pulse neural network model for time-frequency feature fusion according to the classification loss value until the optimal pulse neural network model for time-frequency feature fusion is obtained; The application module is used to detect communication standard data online according to the pulse neural network model fused with time-frequency features, and adaptively identify communication standard data; Among them, the update module is preset with training parameters of the pulse neural network model with time-frequency feature fusion, and the training parameters include: number of training times, batch size, optimizer parameters, learning rate, LIF attenuation factor, LIF reset mechanism and LIF pulse emission threshold.

8. A computer storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

9. An electronic device, characterized in that: It comprises a memory and one or more processors, wherein the memory is used to store one or more programs; when the one or more programs are executed by the one or more processors, the method according to any one of claims 1 to 6 is implemented.

Images