KR101300650B1 - A method of recognizing pri modulation type based on support vector machine, and an apparatus of recognizing pri modulation type - Google Patents

Abstract

The present invention relates to a PRI modulation mode recognition method and a PRI modulation mode recognition method using the same. The PRI modulation mode recognition method includes (a) generating a support vector machine (SVM) for each PRI (Pulse Repetition Interval) modulation type, and Training and storing the generated SVM for each modulation type; (b) extracting feature vectors from time of arrival information of collected pulses for recognizing a PRI modulation type; (c) checking a classification result value by inputting the feature vector extracted in step (b) to the stored SVM; (d) selecting an SVM representing the largest value among the identified classification result values; And (e) recognizing a PRI modulation type using the selected SVM. As a result, the present invention can improve the recognition rate of the PRI modulation type.

Description

A method of recognizing PRI modulation type based on Support Vector Machine, and an apparatus of recognizing PRI modulation type

The present invention relates to a method for recognizing a radar signal pulse repetition interval (PRI) modulation type, and more particularly, to a method for recognizing a PRI modulation type applied to detection of a radar signal of an electronic warfare support system and a PRI modulation using the same. It relates to a shape recognition device.

Electronic warfare support systems, after receiving enemy signals, identify and locate threat emitters to help determine enemy power structures and placement.

In order to identify the radar signal in the electronic warfare support system, the analysis of the frequency, pulse repetition interval (PR), and scan, which are the main factors of the radar signal, must be preceded. Among them, PRI is the most basic element used to identify radar because each radar uses unique modulation type and value.

Methods for analyzing PRI modulation patterns in electronic warfare systems include histogram and autocorrelation, but these methods have some limitations.

The method using the histogram generates the histogram using the difference of TOA (Time Of Arrival) between each pulse and recognizes the PRI modulation type by using the characteristic of the histogram according to each modulation type.

However, this method is difficult to automate due to the ambiguity of histogram bin, threshold, etc. and large sensitivity to signal distortion such as missing pulses and unnecessary signals, so it is not possible to guarantee reliability and accuracy mainly by relying on manual analysis of the operator. Has

The autocorrelation method calculates the autocorrelation between each pulse and defines a type separator for distinguishing each PRI modulation type and applies it to each pulse train. The autocorrelation method should be preceded by filtering that acts as a preprocessor for missing and distorted phenomena. Since the periodicity is used for type classification, a sufficient number of pulses must be collected to confirm the periodicity. There is a disadvantage that the time complexity is high because the relationship is calculated.

KR 10-2010-0121108A, November 17, 2010, Fig. 1

An object of the present invention is to provide a PRI modulation mode recognition method and a PRI modulation mode recognition apparatus using the same that can improve the recognition rate of the PRI modulation type by applying a support vector machine (SVM).

One aspect of the present invention for achieving the above object relates to a PRI (Pulse Repetition Interval) modulation type recognition method applied to the detection of the radar signal of the Electronic Warfare Support (Electronic Warfare Support) system, the PRI modulation mode recognition method (Pulse Repetition Interval) generating a support vector machine (SVM) for each modulation type, and training and storing the generated SVM for each modulation type; (b) extracting feature vectors from time of arrival information of collected pulses for recognizing a PRI modulation type; (c) checking a classification result value by inputting the feature vector extracted in step (b) to the stored SVM; (d) selecting an SVM representing the largest value among the identified classification result values; And (e) recognizing a PRI modulation type using the selected SVM.

Step (a) may include (a1) generating the SVM for each PRI modulation type to be recognized; (a2) generating a plurality of training data consisting of a time of argument (TOA) and a modulation type code for each modulation type; (a3) extracting feature vectors input to the SVM by calculating all TOAs of the training data; (a4) classifying a PRI modulation form value corresponding to a modulation form of the feature vector extracted by step (a3); (a5) training the SVM for each modulation type by receiving transform data consisting of the feature vector extracted in step (a3) and the PRI modulation type values classified in step (a4); And (a6) if there is no training data that did not perform the learning may include the step of storing the training SVM for each modulation type and terminate the training.

D (x), which is the hyperplane of determination of the SVM classifier used in step (c), is represented by Equation 1, and when d (x) is greater than 0, it is classified as the first class, and if it is less than 0, it is classified as the second class. Where w is a weight value, K may be a kernel function, s may be a support vector, and b may be a bias.

Equation 1:

Step (e) may include: (e1) generating a predefined result code according to the number of the selected SVM; And (e2) recognizing the PRI modulation type by comparing the generated result code with a previously defined PRI modulation type table for the result code.

Another aspect of the present invention for achieving the above object relates to a PRI (Pulse Repetition Interval) modulation type recognition device applied to the detection of the radar signal of the Electronic Warfare Support (Electronic Warfare Support) system, the present PRI modulation type recognition device The receiving unit detects and processes an external radar signal; And a controller for performing the PRI modulation type recognition method according to an aspect of the present invention, based on the information detected and processed by the receiver.

As such, the present invention utilizes a decision hyperplane to maximize the margin between classes of each modulation type, thereby reducing the PRI modulation type recognition rate by using a support vector machine (SVM) that minimizes the error rate of pattern recognition. By improving the radar signal detection accuracy, it can be applied to various electronic warfare systems requiring signal analysis.

1 is a flowchart illustrating a method for recognizing a PRI modulation type according to an embodiment of the present invention.
FIG. 2 is a flowchart for describing operation S100 of FIG. 1.
3 is a conceptual diagram of a primary PRI and a secondary PRI according to an embodiment of the present invention.
4 is a block diagram of a PRI modulation type recognition apparatus according to another embodiment of the present invention.

Hereinafter, a radar signal PRI modulation mode recognition method and a PRI modulation mode recognition apparatus according to an embodiment of the present invention will be described with reference to the accompanying drawings.

FIG. 1 is a flowchart illustrating a method for recognizing a PRI modulation type according to an embodiment of the present invention, and FIG. 2 is a flowchart illustrating step S100 of FIG. 1.

The radar signal PRI modulation type recognition method according to the present embodiment uses a decision hyperplane to maximize margins between classes of each modulation type, thereby minimizing an error rate of pattern recognition. Is performed by applying.

SVM is a supervised learning model that analyzes data and recognizes patterns. It is performed using classification and regression analysis.

In the PRI modulation type recognition method according to the present embodiment, as shown in FIG. 1, a step of training a support vector machine (SVM) (S100), extracting a feature vector from input data (S200), and an SVM for each modulation type is performed. Checking the input data classification result value at step S300, selecting the SVM maximizing the classification result value at step S400, generating a result code according to the selected SVM at step S500, and using the result code Recognizing a PRI modulation type may include a step (S600).

Training SVM (Support Vector Machine) (S100), as shown in Figure 2, SVM generation step for each modulation type (S110), training data generation step (S120), training data feature vector extraction step (S130), Training data classification step (S140), SVM learning step (S150), the remaining training data check step (S160) and may end the training step (S170).

The SVM generation step for each modulation type (S110) generates an SVM for each PRI modulation type to be recognized. The training data generation step S120 generates a plurality of training data consisting of a TOA and a modulation type code for each modulation type.

The training data feature vector extraction step (S130) is a step of extracting feature vectors to be input into the SVM from the training data. As shown in FIG. 3, the primary PRI is the difference between the TOAs, and the secondary PRI is the difference between the primary PRIs.

Vector illustration Explanation Probability of increase Probability when the difference between 1st PRI is positive Probability of decrease Probability when the difference between the first PRIs is ?? 1st incremental continuity The probability that the difference between the first PRIs stays positive 2nd incremental continuity Probability that the difference between the 2nd PRI maintains ?? 1st decrease continuity The probability that the difference between the first PRIs remains ?? 2nd Decrease Continuity Probability that the difference between the 2nd PRI maintains ?? Invariant Probability that the difference between the first PRIs is less than or equal to σ _STB (Invariant Threshold)

Training data classification step (S140) is a step of classifying input data input to each SVM. Input data input to the SVM is the feature vector and the PRI modulation type value (0 or 1) extracted in step S130.

Specifically, in the training data classification step S140, when the binary SVM classifier is used in the PRI modulation mode recognition method according to the present embodiment, the PRI modulation type value is set to 1 when inputting to the SVM corresponding to the modulation type of the input data. In other words, when inputting to the SVM corresponding to another modulation type, set the PRI modulation type value to 0 to classify the training data.

SVM learning step (S150) receives the transform data consisting of the feature vector and the PRI modulation type value for the input data to train the SVM for each modulation type.

Residual training data check step (S160) checks whether there is training data that has not performed learning. In step S170, when it is determined in step S160 that there is no training data, the training SVM for each modulation type is stored and the training ends.

By training the SVM (Support Vector Machine) consisting of the steps S110 to S170 as described above (S100) can be implemented SVM trained for each modulation type.

Next, extracting the feature vector from the input data (S200), the feature vector from the collected data (TOA column) for recognizing the PRI modulation form using the same method as the above-described training data feature vector extraction step (S130) Can be extracted.

Next, the classification result value checking step for each SVM (S300) is a step of confirming the classification result value for the input data in the SVM for each modulation type, the SVM for each modulation type implemented in step S100 by using the feature vector extracted in step S200. Enter in to check the classification result.

The crystal hyperplane d (x) of the SVM classifier used in this embodiment may be expressed as Equation 1. If the crystal hyperplane d (x) is greater than 0, it is classified as a first class, and if it is smaller than 0, it is classified as a second class. . In Equation 1, w is a weight value, K is a kernel function, s is a support vector, and b is a bias.

The SVM selection step (S400) is a step of selecting an SVM that maximizes the classification result value, and selects the SVM representing the largest value among the classification result values identified by the step S300.

Result code generation step (S500) generates a predefined result code according to the number of the SVM selected by the step S400. For example, when the third of the classification result value in step S400 has a maximum value, the result code generated in step S500 may be represented by [0010000] as shown in Table 2 below.

The PRI modulation type classification step S600 is a step of recognizing the PRI modulation type using the result code. The PRI modulation type classification step S600 may recognize the PRI modulation type by comparing the result code generated by step S500 with the PRI modulation type table for the predefined result code as shown in Table 2.

Result code PRI modulation type 1000000 Dwell & Switch 0100000 Wobble 0010000 Jitter 0001000 Linear sliding (+) 0000100 Linear sliding (-) 0000010 Nonlinear sliding (+) 0000001 Nonlinear sliding (-)

Hereinafter, the PRI modulation mode recognition apparatus 1 according to an embodiment of the present invention will be applied to a radar signal detection apparatus for detecting and identifying a radar signal, with reference to FIG. 4. The method can be used to operate.

As shown in FIG. 4, the PRI modulation mode recognition device 1 according to the present embodiment may include a receiver 10 for detecting a radar signal and a controller 20 performing the above-described PRI modulation mode recognition method. have.

The receiver 10 detects the radar signal, processes the detected signal, and transmits the detected signal to the controller 20.

When the radar signal is detected, the receiver 10 may measure a carrier frequency pulse width, a pulse amplitude, a time of arrival of the signal, and the like.

The control unit 20 is functionally SVM training module 22, input data feature vector extraction module 24, SVM classification result value confirmation module 26 as shown in Figure 3 to implement the PRI modulation type recognition method described above ) And PRI modulation type classification module 28. This functional division is a division for convenience and the control unit 20 according to the present invention is not necessarily made of the above functional blocks.

The SVM training module 22 is a module for performing the control procedure according to the flowchart of FIG. 2 described above. The SVM training module 22 generates an SVM for each PRI modulation type, and generates a plurality of training data consisting of a TOA and a modulation type code for each modulation type. do.

The SVM training module 22 extracts a feature vector input to the SVM from the generated training data, classifies the input data input to the SVM including the extracted feature vector and the PRI modulation type value, and uses each input data to input the feature vector. Train SVMs by modulation type.

The SVM training module 22 checks whether there is training data that has not been learned, and stores and terminates the SVM trained for each modulation type.

The input data feature vector extraction module 24 extracts the feature vector from the collection data (TOA column) for recognizing the PRI modulation form using the same method as the above-described training data feature vector extraction step according to step S130 of FIG. 2. Can be.

The classification result value checking module 26 for each SVM checks the classification result value for the input data in the SVM for each modulation type and extracts the feature vectors extracted by the input data feature vector extraction module 24 for the SVM training module 22. The result of classification is put into SVM for each modulation type implemented by.

The PRI modulation type classification module 28 selects the SVM representing the largest value among the identified classification results, generates a result code according to the number of the selected SVM, and recognizes the PRI modulation type by using the result code. can do.

As described above, the PRI modulation mode recognition method and the PRI modulation mode recognition apparatus 1 according to an embodiment of the present invention extract a feature vector using TOA information of collected pulses, and use PRI to modulate a PRI using an SVM. By recognizing the shape, the modulation type recognition rate and the recognition speed can be improved.

1: PRI modulation type recognition device
10: receiver
20: control unit
22: SVM Training Module
24: input data feature vector extraction module
26: Classification result value checking module for each SVM
28: PRI modulation type classification module

Claims (5)

In the PRI (Pulse Repetition Interval) modulation type recognition method applied to the detection of the radar signal of the electronic warfare support system,
(a) generating a support vector machine (SVM) for each pulse repetition interval (PRI) modulation type, and training and storing the generated SVM for each modulation type;
(b) extracting feature vectors from time of arrival information of collected pulses for recognizing a PRI modulation type;
(c) checking a classification result value by inputting the feature vector extracted in step (b) to the stored SVM;
(d) selecting an SVM representing the largest value among the identified classification result values; And
(e) recognizing a PRI modulation type using the selected SVM. The method of claim 1,
The step (a)
(a1) generating the SVM for each PRI modulation type to be recognized;
(a2) generating a plurality of training data consisting of a time of argument (TOA) and a modulation type code for each modulation type;
(a3) extracting feature vectors input to the SVM by calculating all TOAs of the training data;
(a4) classifying a PRI modulation form value corresponding to a modulation form of the feature vector extracted by step (a3);
(a5) training the SVM for each modulation type by receiving transform data consisting of the feature vector extracted in step (a3) and the PRI modulation type values classified in step (a4); And
(a6) if there is no training data that has not been trained, storing the SVM trained for each modulation type and ending the training. The method of claim 1,
D (x), the hyperplane of determination of the SVM classifier used in step (c), is represented by Equation 1, and when d (x) is greater than 0, it is classified as the first class, and if it is less than 0, it is classified as the second class. In 1, w is a weight value, K is a kernel function, s is a support vector, and b is a bias.
Equation 1:

The method of claim 1, wherein step (e)
(e1) generating a predefined result code according to the number of the selected SVM; And
(e2) recognizing the PRI modulation form by comparing the generated result code with a table of PRI modulation forms for the result code defined in advance. In the PRI (Pulse Repetition Interval) modulation type recognition device applied to the detection of the radar signal of the Electronic Warfare Support system,
A receiver for detecting and processing an external radar signal; And
And a controller for performing the PRI modulation mode recognition method according to any one of claims 1 to 4, based on the information detected and processed by the receiving unit.

Images