Civil aviation communication navigation broadband electromagnetic interference source identification method

Abstract

The present application relates to the technical field of interference source identification, in particular to a civil aviation communication navigation wideband electromagnetic interference source identification method, comprising the following steps: step one, data preprocessing is performed on original frequency domain data; step two, feature extraction is performed on the processed data; step three, relief-F feature screening and random forest algorithm parameter joint optimization: this part is mainly divided into constructing an objective function and self-adaptive particle swarm optimization; the civil aviation communication navigation wideband electromagnetic interference source identification method firstly extracts the features of the preprocessed interference source spectrum data, then uses the relief-F algorithm for further feature screening, and finally realizes interference source classification and identification by using the random forest algorithm; according to the different characteristics of the interference source signals, eight types of features that can accurately represent the characteristics of different interference sources are selected, and the improved PSO algorithm is used to jointly optimize the relief-F algorithm parameter screening threshold and the random forest algorithm parameter decision tree number n_estimators and the maximum feature number max_features, so as to improve the accuracy of the identification result.

Description

Technical Field

[0001] This invention relates to the field of interference source identification technology, specifically a method for identifying broadband electromagnetic interference sources in civil aviation communication and navigation. Background Technology

[0002] As is well known, with the increasing variety and quantity of electronic and electrical equipment around airports, the threat of electromagnetic interference to civil aviation communication and navigation systems is also rising. The diverse broadband electromagnetic interference sources around civil airports, due to their strong randomness and weak regularity, make it difficult to determine the scope and direction of investigation, thus increasing the difficulty of investigation. Furthermore, with the continuous improvement and development of integrated transportation systems, high-speed trains are passing near airports, and the electromagnetic radiation generated when these trains pass through the electrical phase-splitting area may also cause potential interference to the airport's communication and navigation signals.

[0003] Numerous scholars have already tested and analyzed the electromagnetic environment of airports, providing an important foundation for the study of electromagnetic interference sources. However, research on the characteristic analysis and identification of broadband electromagnetic interference sources for civil aviation communication and navigation is still limited, especially research on the identification of sources in the presence of overlapping interference. This patent proposes a method based on optimized feature selection and random forest to identify broadband electromagnetic interference sources for civil aviation communication and navigation. This method can effectively and quickly identify different types of electromagnetic interference sources, thereby ensuring the safety of civil aviation communication and navigation. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the shortcomings of existing technologies, this invention provides a method for identifying broadband electromagnetic interference sources in civil aviation communication and navigation.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying broadband electromagnetic interference sources in civil aviation communication and navigation, comprising the following steps:

[0008] Step 1: Perform data preprocessing on the raw frequency domain data;

[0009] Step 2: Extract features from the processed data;

[0010] Step 3: Joint optimization of Relief-F feature selection and random forest parameters: This part mainly consists of constructing the objective function and adaptive particle swarm optimization;

[0011] Step 4: Feature Filtering: The original feature dataset is filtered using the Relief-F algorithm with an optimal filtering threshold. The filtered feature data is used as the input feature dataset for the random forest model.

[0012] Step 5: Random Forest Classification and Recognition: Based on the filtered feature dataset, first train the optimized random forest model, and then use the trained random forest model to classify and identify interference sources.

[0013] Preferably, in step one, data preprocessing involves performing moving average noise reduction on the original frequency domain data.

[0014] Preferably, in step two, the specific feature extraction includes: first, dividing the preprocessed interference source signal frequency domain data into ten segments according to frequency bands, and then extracting the maximum value, minimum value, root mean square, waveform factor, skewness, centroid frequency, root mean square frequency, and frequency standard deviation for each segment of frequency domain data. The feature data obtained from feature extraction is the original feature dataset.

[0015] Preferably, step three specifically includes the following steps:

[0016] The first step is to construct an objective function for the joint optimization of Relief-F feature selection and random forest parameters using the Adaptive Particle Swarm Optimization (APSO) algorithm. Step 1: The Relief-F algorithm parameter selection threshold δ, and the random forest parameters (number of decision trees n_estimators and maximum number of features max_features) are used as optimization variables. Then, the Relief-F algorithm and the random forest algorithm are embedded within the objective function. Step 2: The obtained original feature dataset is used as input to the Relief-F feature selection algorithm. The Relief-F algorithm selects the original feature dataset according to the selection threshold δ. Finally, the 3-fold cross-validation result of the random forest algorithm on the selected feature dataset under the parameter settings of the number of decision trees n_estimators and the maximum number of features max_features is used as the objective function value corresponding to the current optimization variable.

[0017] The second step is to use the APSO algorithm to optimize the screening threshold δ and n_estimators, max_features. The main implementation method is: in each iteration, the particles are sorted from best to worst according to their individual optimal positions, and different inertial weights and acceleration coefficients are assigned to different particles. The inertial weight, acceleration coefficient, particle velocity and position of the i-th particle (i = 1, 2, 3, ..., N, where N is the population size) are calculated as shown in equations (1) to (4). When APSO completes the optimization, it will output the best value of the optimization variables.

[0018]

[0019]

[0020]

[0021]

[0022] Among them, w i w is the inertial weight of the i-th particle. max w min These are the predefined maximum and minimum inertia weights, with a default value of w. max =0.9, take w min =0.55; c i1 c i2 V is the acceleration coefficient of the i-th particle; k is the iteration number; V id X represents the d-th dimension of the velocity of the i-th particle. id Let P be the d-th dimension of the position of the i-th particle; id-pbest P represents the d-th dimension of the optimal position of the i-th individual particle. d-gbest r is the d-th dimension of the global optimal position of the particle swarm; 1d r 2d It is a random number uniformly distributed in the range [0,1].

[0023] (III) Beneficial Effects

[0024] Compared with existing technologies, this invention provides a method for identifying broadband electromagnetic interference sources in civil aviation communication and navigation, which has the following advantages:

[0025] This method for identifying broadband electromagnetic interference sources in civil aviation communication and navigation first extracts the preprocessed spectral data features of the interference sources, then uses the Relief-F algorithm for further feature filtering, and finally utilizes the Random Forest algorithm to classify and identify the interference sources. Based on the different characteristics of the interference source signals, eight types of features that can accurately characterize the characteristics of different interference sources are selected. To further improve the accuracy and efficiency of identification, an improved PSO algorithm is used to jointly optimize the filtering threshold of the Relief-F algorithm parameters and the number of decision trees n_estimators and the maximum number of features max_features of the Random Forest algorithm parameters, thereby improving the accuracy of the identification results. Attached Figure Description

[0026] Figure 1 This is a diagram illustrating the overall framework of the interference source identification model for the present invention.

[0027] Figure 2 This is a comparison of the waveforms of the six-wire shield of the present invention and the interference signals of a household light bulb after noise reduction;

[0028] Figure 3 This is a diagram showing the weight distribution of structural features in this invention.

[0029] Figure 4 This is a confusion matrix diagram of the classification and recognition results of the random forest algorithm of this invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Please see Figure 1-4 A method for identifying broadband electromagnetic interference sources in civil aviation communication and navigation includes the following steps:

[0032] Step 1: Perform data preprocessing on the raw frequency domain data;

[0033] Step 2: Extract features from the processed data;

[0034] Step 3: Joint optimization of Relief-F feature selection and random forest parameters: This part mainly consists of constructing the objective function and adaptive particle swarm optimization;

[0035] Step 4: Feature Filtering: The original feature dataset is filtered using the Relief-F algorithm with an optimal filtering threshold. The filtered feature data is used as the input feature dataset for the random forest model.

[0036] Step 5: Random Forest Classification and Recognition: Based on the filtered feature dataset, first train the optimized random forest model, and then use the trained random forest model to classify and identify interference sources.

[0037] To facilitate the elimination of interference, in this embodiment, step one, data preprocessing, involves performing moving average noise reduction on the original frequency domain data.

[0038] To facilitate the extraction of different features and ensure the accuracy of the results, in this embodiment, the specific feature extraction in step two includes: firstly, dividing the preprocessed interference source signal frequency domain data into ten segments according to frequency bands, and then extracting the maximum value, minimum value, root mean square, waveform factor, skewness, centroid frequency, root mean square frequency, and frequency standard deviation for each segment of frequency domain data. The feature data obtained from feature extraction is the original feature dataset.

[0039] The specific steps in step three include:

[0040] The first step is to construct the objective function for the joint optimization of Relief-F feature selection and random forest parameters using the APSO algorithm. Step 1: The Relief-F algorithm parameter selection threshold δ, and the random forest parameters (number of decision trees n_estimators and maximum number of features max_features) are used as optimization variables. Then, the Relief-F algorithm and the random forest algorithm are embedded into the objective function. Step 2: The obtained original feature dataset is used as the input to the Relief-F feature selection algorithm. The Relief-F algorithm selects the original feature dataset according to the selection threshold δ. Finally, the 3-fold cross-validation result of the random forest algorithm on the selected feature dataset under the parameter settings of the number of decision trees n_estimators and the maximum number of features max_features is used as the objective function value corresponding to the current optimization variable.

[0041] The second step is to use the APSO algorithm to optimize the screening threshold δ and n_estimators, max_features. The main implementation method is: in each iteration, the particles are sorted from best to worst according to their individual optimal positions, and different inertial weights and acceleration coefficients are assigned to different particles. The inertial weight, acceleration coefficient, particle velocity and position of the i-th particle (i = 1, 2, 3, ..., N, where N is the population size) are calculated as shown in equations (1) to (4). When APSO completes the optimization, it will output the best value of the optimization variables.

[0042]

[0043]

[0044]

[0045]

[0046] Among them, w i w is the inertial weight of the i-th particle. max w min These are the predefined maximum and minimum inertia weights, with a default value of w. max =0.9, take w min =0.55; c i1 c i2 V is the acceleration coefficient of the i-th particle; k is the iteration number; V id X represents the d-th dimension of the velocity of the i-th particle. id Let P be the d-th dimension of the position of the i-th particle; id-pbest P represents the d-th dimension of the optimal position of the i-th individual particle. d-gbest r is the d-th dimension of the global optimal position of the particle swarm; 1d r 2dIt is a random number uniformly distributed in the range [0,1].

[0047] Example 1:

[0048] The interference sources used in this example include electromagnetic radiation from trains passing through electrical phase-splitting points, three types of aging light bulbs (airport runway bulbs, airport lighting bulbs, and civilian bulbs), two types of signal jammers (two-wire shields and six-wire shields), and eight aliased interference sources formed by the superposition of the above six types of interference sources, totaling 14 types of interference sources. Electromagnetic radiation data from trains passing through electrical phase-splitting points were measured on a railway line in Chengdu, Sichuan Province. Data from the other interference sources were measured in a laboratory. A total of 16 sets of electromagnetic radiation samples from trains passing through electrical phase-splitting points were collected, and 19 sets of samples were collected from each of the other interference sources.

[0049] Step 1: Perform moving average denoising on the original frequency domain data of the interference source signal. Compare the denoised waveforms of the original signals from the household light bulb and the six-wire shielded interference source. Figure 2 As shown.

[0050] Step 2: First, the denoised data is evenly divided into ten segments according to frequency to facilitate subsequent feature extraction. Based on the analysis of the waveforms and characteristics of different interference sources, the following features are selected for extraction: maximum value, minimum value, root mean square (RMS), waveform factor, skewness, centroid frequency, frequency standard deviation, and RMS frequency. Using the selected eight features, eight categories of features are extracted from the segmented data, resulting in an original dataset with 80 features for each sample.

[0051] Step 3 involves using the APSO algorithm to find the optimal feature selection parameters for the Relief-F algorithm (screening threshold δ) and the Random Forest algorithm (n_estimators and max_features). First, an objective function is constructed, using the Relief-F algorithm's screening threshold δ and the Random Forest parameters n_estimators and max_features as optimization variables. Then, the Relief-F algorithm and the Random Forest algorithm are embedded into the APSO algorithm's objective function. The original feature dataset obtained in Step 2 is used as the input to the Relief-F algorithm within the objective function. The Relief-F algorithm filters the original feature dataset based on the screening threshold δ. Finally, the 3-fold cross-validation result of the Random Forest algorithm on the filtered feature dataset, with the number of decision trees n_estimators and the maximum number of features max_features set, is used as the objective function value corresponding to the current optimization variables. At the end of the search, APSO returns the screening threshold δ and n_estimators and max_features.

[0052] In Example 1, the optimal screening threshold δ was 0.078, the random forest parameter n_estimators was 93, and the max_features was 11.

[0053] Step 4: Set the feature selection threshold δ = 0.078 obtained in Step 3 into the Relief-F algorithm, and then perform feature selection on the 80-dimensional original dataset obtained in Step 2. In Example 1, the weight distribution of the 80-dimensional features extracted in Step 2 is as follows: Figure 3 As shown in the figure. After filtering, a 70-dimensional filtered feature dataset was obtained.

[0054] Step 5: Set the optimal parameter values ​​obtained in Step 3 (n_estimators = 93, max_features = 11) into the random forest algorithm. Use the 70-dimensional filtered feature dataset obtained in Step 4 as input to identify 14 types of interference source signals. The identification accuracy is 96.4%, and the confusion matrix of the classification result is as follows. Figure 4 As shown:

[0055] exist Figure 4 In the confusion matrix shown, the A to N categories on the horizontal and vertical axes represent the 14 types of interference sources used in Example 1. These are two types of signal jammers (two-wire shields and six-wire shields), electromagnetic radiation from trains passing through the electrical phase-splitting circuit, three types of aging light bulbs (airport runway bulbs, airport lighting bulbs, and civilian bulbs), and eight types of aliased interference sources formed by the aliasing of the above six types of interference sources: aliasing of two-wire shields with airport lighting bulbs, aliasing of two-wire shields with civilian bulbs, aliasing of six-wire shields with airport runway bulbs, aliasing of six-wire shields with airport lighting bulbs, aliasing of six-wire shields with civilian bulbs, aliasing of two-wire shields with six-wire shields, aliasing of airport runway bulbs with airport lighting bulbs, and aliasing of airport lighting bulbs with civilian bulbs. Figure 4 It can be seen that the interference source identification method provided by the present invention has an accuracy rate of 100% for identifying the 11 types of interference sources mentioned above, and an accuracy rate of 83% for identifying interference sources such as six-wire shields mixed with airport runway light bulbs, six-wire shields mixed with civilian light bulbs, and airport lighting bulbs mixed with civilian light bulbs.

[0056] To illustrate the possible application scenarios, technical principles, implementable specific solutions, and achievable objectives and effects of this application in detail, the following description, in conjunction with the listed specific embodiments and accompanying drawings, provides a detailed explanation. The embodiments described herein are merely illustrative of the technical solutions of this application and are therefore intended to limit the scope of protection of this application.

[0057] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The term "embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor does it specifically limit its independence or connection with other embodiments. In principle, in this application, as long as there are no technical contradictions or conflicts, the technical features mentioned in each embodiment can be combined in any way to form corresponding implementable technical solutions.

[0058] Unless otherwise defined, the technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the use of related terms herein is merely for the purpose of describing particular embodiments and is not intended to limit this application.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for identifying broadband electromagnetic interference sources in civil aviation communication and navigation, characterized in that, Includes the following steps: Step 1: Perform data preprocessing on the raw frequency domain data; Step 2: Extract features from the processed data; Step 3: Joint optimization of Relief-F feature selection and random forest algorithm parameters: This part mainly consists of constructing the objective function and adaptive particle swarm optimization; Step 4: Feature Filtering: The Relief-F algorithm with an optimal filtering threshold is used to filter the features in the original feature dataset. The filtered feature data is used as the input feature dataset for the random forest model. Step 5: Random Forest Classification and Recognition: Based on the filtered feature dataset, first train the optimized random forest model, and then use the trained random forest model to classify and identify interference sources. The specific steps in step three include: The first step is to construct an objective function for the joint optimization of Relief-F feature selection and random forest algorithm parameters using the Adaptive Particle Swarm Optimization (APSO) algorithm. Step 1 involves using the Relief-F algorithm parameter selection threshold δ and the random forest parameters (number of decision trees n_estimators and maximum number of features max_features) as optimization variables. Then, the Relief-F algorithm and the random forest algorithm are embedded within the objective function. Step 2 uses the obtained original feature dataset as input to the Relief-F feature selection algorithm. The Relief-F algorithm selects the original feature dataset based on the selection threshold δ. Finally, the 3-fold cross-validation result of the random forest algorithm on the selected feature dataset under the parameter settings of the number of decision trees n_estimators and the maximum number of features max_features is used as the objective function value corresponding to the current optimization variable. The second step is to use the APSO algorithm to optimize the screening threshold δ and n_estimators, max_features. The main implementation method is: in each iteration, the particles are sorted from best to worst according to their individual optimal positions, and different inertial weights and acceleration coefficients are assigned to different particles. The inertial weight, acceleration coefficient, particle velocity and position calculation method of the i-th particle (i=1, 2, 3, ..., N, N is the population size) are shown in equations (1) to (4). When APSO completes the optimization, it will output the best value of the optimization variables. Where wi is the inertia weight of the i-th particle, wmax and wmin are the predefined maximum and minimum inertia weights, respectively, with default values ​​of wmax = 0.9 and wmin = 0.55; ci1 and ci2 are the acceleration coefficients of the i-th particle; k is the number of iterations; Vid is the d-th dimension of the velocity of the i-th particle; Xid is the d-th dimension of the position of the i-th particle; Pid-pbest is the d-th dimension of the optimal position of the i-th particle; Pd-gbest is the d-th dimension of the global optimal position of the particle swarm; r1d and r2d are uniformly distributed random numbers in the range [0,1].

2. The method for identifying broadband electromagnetic interference sources in civil aviation communication and navigation according to claim 1, characterized in that: Step one, data preprocessing, involves performing moving average noise reduction on the original frequency domain data.

3. The method for identifying broadband electromagnetic interference sources in civil aviation communication and navigation according to claim 2, characterized in that: In step two, the specific feature extraction includes: first, dividing the preprocessed interference source signal frequency domain data into ten segments according to frequency bands; then, extracting the maximum value, minimum value, root mean square, waveform factor, skewness, centroid frequency, root mean square frequency, and frequency standard deviation for each segment of frequency domain data; the feature data obtained from feature extraction is the original feature dataset.

Images