Simulation generation method of radar signal data set for deep learning

Abstract

The invention provides a simulation generation method of a radar signal data set for deep learning, belongs to the technical field of radar signal preprocessing, and solves the problems that traditional artificial simulation radar signal acquisition is limited by equipment, environment, safety and the like, the acquisition scene is single and the acquisition cost is high. Comprising the following steps: determining basic parameters of a radar signal in an MATLAB environment; determining the type of a modulation mode required to be included in the radar signal data set according to a setting requirement; parameters of different modulation modes of radar signals are set, radar signals corresponding to the modulation modes are generated, white Gaussian noise is added to the radar signals of various modulation modes, and different signal-to-noise ratio parameters are set to enrich a radar signal data set; respectively storing the processed radar signal data into corresponding text files in the form of IQ data; and reading the generated text data in a python environment, outputting files in a pkl format, and integrating all the files as a final radar signal data set.

Description

Technical Field

[0001] The present invention relates to a method for simulating and generating a radar signal data set for deep learning, belonging to the technical field of radar signal preprocessing. Background Art

[0002] Radar signal data sets have important research value and application significance in modern military, civil aviation, meteorological monitoring and other fields. However, obtaining real radar signal data faces many challenges and limitations, which makes artificial simulation of radar signals particularly important in practical research.

[0003] From the perspective of the background art, the acquisition of radar signal data mainly relies on the collection of actual radar systems. This collection process faces multiple difficulties: First, radar equipment is expensive, with high operation and maintenance costs, and requires professional operators and perfect technical support; Second, the collection of radar signals is often restricted by various factors such as geographical location, weather conditions, and electromagnetic environment, making it difficult to ensure the continuity and integrity of data collection.

[0004] In this context, artificial simulation of radar signals has become a necessary supplementary means, with the following characteristics and deficiencies:

[0005] 1. Limited data acquisition: The acquisition of real radar signal data is often restricted by multiple factors such as equipment, environment, and security, resulting in a serious shortage of available research data. Especially in machine learning applications that require a large number of training samples, data scarcity has become an important factor restricting technological development.

[0006] 2. Single scenario: The actually collected radar data is often limited to specific geographical locations and application scenarios, and it is difficult to cover various complex situations, such as signal characteristics in special environments such as extreme weather conditions, multi-target interference, and electronic countermeasures.

[0007] 3. High acquisition cost: Building and maintaining a radar system requires a huge capital investment, and the consumption of manpower and material resources during the data collection process is also quite considerable, making it difficult to carry out large-scale measured data collection activities.

[0008] 4. Data security sensitivity: Signal data in fields such as military radar and civil aviation often involves sensitive information and requires strict confidentiality measures, which greatly restricts the circulation and sharing of data.

[0009] 5. Influence of environmental factors: Actual radar signals are easily affected by environmental factors such as weather, terrain, and electromagnetic interference, making it difficult to ensure the stability and reliability of data quality.

[0010] Based on the above difficulties, artificial simulation of radar signals has become an important alternative. Through computer simulation, the characteristics of radar signals in various complex scenarios can be simulated, generating a large amount of training data to support algorithm research and performance evaluation. However, there are also differences between simulation data and real signals. How to improve the authenticity and reliability of simulation data is an issue that needs to be focused on in current research.

[0011] Therefore, in practical applications, it is often necessary to combine limited real data with simulation data to ensure both the authenticity of the data and meet the research's requirements for the amount of data. Although this method cannot completely solve the difficulty of data acquisition, it provides a feasible approach for the research and development of radar signal processing technology. Summary of the Invention

[0012] The present invention aims to solve the problems that the acquisition of traditional artificial simulation radar signals is restricted by multiple factors such as equipment, environment, and safety, the acquisition scenarios are single, and the acquisition cost is high. Furthermore, a method for simulating and generating a radar signal dataset for deep learning is proposed.

[0013] The technical solution adopted by the present invention to solve the above problems is as follows: The present invention includes the following steps:

[0014] Step 1: Determine the basic parameters of the radar signal in the MATLAB environment;

[0015] Step 2: According to the set requirements, determine the types of modulation methods to be included in the radar signal dataset;

[0016] Step 3: Set the parameters of different modulation methods of the radar signal, generate radar signals corresponding to the modulation methods, add white Gaussian noise to the radar signals of various modulation methods, and set different signal-to-noise ratio parameters to enrich the radar signal dataset, and complete the processing of the radar signals corresponding to the modulation methods;

[0017] Step 4: Store the processed radar signal data in the form of IQ data into the corresponding text files respectively;

[0018] Step 5: Read the generated text data in the python environment, output files in pkl format, and integrate all files as the final radar signal dataset.

[0019] Furthermore, the basic parameters of the radar signal in Step 1 include at least the sampling rate, sampling time, center frequency, and amplitude of the radar signal.

[0020] Furthermore, the types of modulation methods in Step 2 include LFM, NLFM, Barker code Costas loop, conventional pulse signal, P1, P2, P3, and P4.

[0021] Further, step 3 specifically includes:

[0022] Step 3.1: Set the required parameters for the corresponding modulation method to generate a radar signal of the corresponding modulation method;

[0023] Step 3.2: Initialize the generated radar signal. For different modulation methods, set the signal-to-noise ratio parameter of the corresponding basic parameters of the radar signal, and add white Gaussian noise to the initialized radar signal according to the corresponding signal-to-noise ratio parameter;

[0024] Step 3.3: Normalize the radar signal after adding noise;

[0025] Step 3.4: Extract the real part and the imaginary part of the normalized radar signal, and transfer the extracted real part and imaginary part to the save function.

[0026] Further, in step 3.2, the signal-to-noise ratio parameter is set to -20 dB to 10 dB, with a step of 2 dB.

[0027] Further, step 4 specifically includes:

[0028] Step 4.1: Intercept the signal part in the real part and the imaginary part;

[0029] Step 4.2: Convert the signal-to-noise ratio parameter of the corresponding modulation model into a string, and name the txt text file with the string;

[0030] Step 4.3: Store the intercepted signal part of the real part and the imaginary part into the corresponding txt text file.

[0031] Further, step 5 specifically includes:

[0032] Step 5.1: Read the generated text data in the python environment;

[0033] Step 5.2: Clip the read text data and label it according to the modulation method and the signal ratio parameter used by the radar signal;

[0034] Step 5.3: Output the labeled radar signal data as a file in pkl format, and integrate all files into the final radar signal data set.

[0035] The beneficial effects of the present invention are:

[0036] (1) The radar signal data set provided by the present invention includes 8 modulation methods, and the selected modulation methods are all relatively typical and commonly used. In addition, the signal-to-noise ratio range is reasonably set, and it can be directly used for the training and effect analysis of neural networks.

[0037] (2) The radar data set provided by the present invention has scalability. On the basis of the original data, new modulation methods can be added according to requirements, and the signal-to-noise ratio range and various parameter implementation methods that can be used can be modified. The method is simple and has high reusability, solving the problems of limited acquisition of traditional artificial simulation radar signals, single acquisition scenarios, and high acquisition costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flowchart of a method for simulating and generating a radar signal data set for deep learning provided by the present invention;

[0039] Figure 2 It is a flowchart of the modulation method of the chirp signal provided by the present invention;

[0040] Figure 3 It is a schematic diagram of the storage of the modulated radar signal provided by the present invention;

[0041] Figure 4 It is a flowchart of reading text and integrating it into the final radar signal data set provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] Combined with Figures 1-4 to illustrate this embodiment. As Figure 1 shown, the steps of a method for simulating and generating a radar signal data set for deep learning described in this embodiment include:

[0043] S1: Determine the basic parameters of the radar signal in the MATLAB environment;

[0044] In this embodiment, basic parameters such as the sampling rate, sampling time, center frequency, and amplitude of the radar signal need to be determined.

[0045] S2: Determine the types of modulation methods to be included in the radar signal data set according to the setting requirements;

[0046] In this embodiment, 8 modulation methods are included in this data set, namely: LFM, NLFM, Barker code Costas loop, conventional pulse signal, P1, P2, P3, and P4. The radar signal data set provided by the present invention includes 8 modulation methods, and the selected modulation methods are all typical and commonly used. In addition, the signal-to-noise ratio range is set reasonably and can be directly used for the training and effect analysis of neural networks.

[0047] S3: Set the parameters of different modulation methods of the radar signal, generate the radar signal corresponding to the modulation method, add white Gaussian noise to the radar signal of various modulation methods, and set different signal-to-noise ratio parameters to enrich the radar signal data set to complete the processing of the radar signal corresponding to the modulation method;

[0048] AsFigure 2 As shown in the figure, this embodiment takes LFM (Linear Frequency Modulation signal) as an example for illustration. Other modulation methods are similar to LFM and include the following steps:

[0049] S301: Set the required parameters for the corresponding modulation method and generate a radar signal of the corresponding modulation method;

[0050] S302: Generate a zero matrix of the noise-added signal;

[0051] S303: Initialize the generated radar signal. For different modulation methods, set the signal-to-noise ratio parameter of the basic parameters of the corresponding radar signal, and add white Gaussian noise to the initialized radar signal according to the corresponding signal-to-noise ratio parameter;

[0052] S304: Normalize the radar signal after adding noise;

[0053] S305: Extract the real part and the imaginary part of the normalized radar signal;

[0054] S306: Pass the extracted real part and imaginary part to the save function.

[0055] In summary, the radar data set provided by the present invention has scalability. On the basis of the original data, new modulation methods can be added according to requirements, and the signal-to-noise ratio range and various parameter implementation methods that can be used therein can be modified. The method is simple and has high reusability, solving the problems of limited collection of traditional artificial simulation radar signals, single acquisition scenario, and high acquisition cost.

[0056] S4: Store the processed radar signal data in the corresponding text files in the form of IQ data;

[0057] As Figure 3 shown, the steps for saving radar signal data include:

[0058] S401: Intercept the signal parts in the real part and the imaginary part;

[0059] S402: Convert the signal-to-noise ratio parameter of the corresponding modulation model into a string and name the txt text file with the string;

[0060] S403: Store the intercepted signal parts in the real part and the imaginary part into the corresponding txt text file.

[0061] S5: Read the generated text data in the python environment, output a file in pkl format, and integrate all files as the final radar signal data set;

[0062] As Figure 4 shown, the steps for reading the text and integrating it into the final radar signal data set include:

[0063] S501: Read the generated text data in the python environment;

[0064] S502: Clip the read text data and label it according to the modulation method used by the radar signal and the signal ratio parameter;

[0065] S503: Output the labeled radar signal data as a file in pkl format and integrate all files into the final radar signal data set.

[0066] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments with equivalent changes by using the above-disclosed technical content within the scope of the technical solution of the present invention. However, as long as it does not depart from the content of the technical solution of the present invention, any simple modification, equivalent replacement, and improvement made to the above embodiments according to the technical essence of the present invention within the spirit and principle of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for simulating and generating a radar signal data set for deep learning, characterized in that: The steps of the simulation generation method of a radar signal data set for deep learning include: Step 1: Determine the basic parameters of the radar signal in the MATLAB environment; Step 2: Determine the types of modulation methods that need to be included in the radar signal dataset based on the setting requirements; Step 3: Set the parameters of different modulation modes of radar signals, generate radar signals of corresponding modulation modes, add white Gaussian noise to radar signals of various modulation modes, and set different signal-to-noise ratio parameters to enrich the radar signal data set, thus completing the processing of radar signals of corresponding modulation modes; Step 4: Store the processed radar signal data in the form of IQ data into corresponding text files; Step 5: Read the generated text data in the Python environment, output the pkl format file, and integrate all the files as the final radar signal dataset.

2. The method for simulating and generating a radar signal data set for deep learning according to claim 1, characterized in that: The basic parameters of the radar signal in step 1 include at least the sampling rate, sampling time, center frequency and amplitude of the radar signal.

3. The method for simulating and generating a radar signal data set for deep learning according to claim 1, characterized in that: The types of modulation in step 2 include LFM, NLFM, Barker code Costa loop, normal pulse signal, P1, P2, P3 and P4.

4. The method for simulating and generating a radar signal data set for deep learning according to claim 1, characterized in that: Step 3 specifically includes: Step 3.1: Set the required parameters of the corresponding modulation mode and generate the radar signal of the corresponding modulation mode; Step 3.2: Initialize the generated radar signal, set the signal-to-noise ratio parameters of the basic parameters of the corresponding radar signal for different modulation modes, and add white Gaussian noise to the initialized radar signal according to the corresponding signal-to-noise ratio parameters; Step 3.3: Normalize the radar signal after adding noise; Step 3.4: Extract the real and imaginary parts of the normalized radar signal and pass the extracted real and imaginary parts to the save function.

5. The method for simulating and generating a radar signal data set for deep learning according to claim 4, characterized in that: In step 3.2, the signal-to-noise ratio parameter is set to -20dB to 10dB, with a step of 2dB.

6. The method for simulating and generating a radar signal data set for deep learning according to claim 1, characterized in that: Step 4 specifically includes: Step 4.1: Cut off the signal part in the real part and the imaginary part; Step 4.2: Convert the signal-to-noise ratio parameter of the corresponding modulation model into a string and name the txt text file with the string; Step 4.3: Store the intercepted real and imaginary signal parts into the corresponding txt text files.

7. The method for simulating and generating a radar signal data set for deep learning according to claim 1, characterized in that: Step 5 specifically includes: Step 5.1: Read the generated text data in the Python environment; Step 5.2: Cut the read text data and label it according to the modulation method and signal ratio parameters used by the radar signal; Step 5.3: Output the labeled radar signal data as a pkl format file and integrate all files into the final radar signal dataset.

Images