KR101313051B1 - Method and device for recognizing pulse repetition interval modulation type of a radar signal - Google Patents

Abstract

PURPOSE: A pulse repetition interval (PRI) modulation form recognizing method and a recognition apparatus applying the same are provided to easily automatize by simplifying an algorithm and to have low temporal complexity. CONSTITUTION: A PRI modulation form recognizing apparatus (200) comprises a chain code acquiring unit (210), a state transition probability matrix acquiring unit (220), a PRI modulation form separation factor acquiring unit (230), and a modulation form determining unit (240). The code acquiring unit obtains a chain code for an input pulse row. The state transition probability matrix acquiring unit obtains a state transition probability matrix based on the chain code. The PRI modulation form separation factor acquiring unit obtains a PRI modulation form separation factor based on the state transition probability matrix. The modulation form determining unit obtains an output result for a predetermined neural network based on the state transition probability matrix and the modulation form separation factor and determines a PRI modulation form based on the output result. [Reference numerals] (110) Signal receiving unit; (120) Signal processing unit; (200) Modulation form recognizing apparatus; (210) Chain code acquiring unit; (220) State transition probability matrix acquiring unit; (230) PRI modulation form separation factor acquiring unit; (240) Modulation form determining unit; (250) Nerve network discipline unit; (260) Nerve network processing unit

Description

Pulse Repetition Period Modulation Type Recognition Method and Recognition Apparatus Applying the Same

The present invention relates to a neural network based PRI modulation type recognition method using a state transition probability and a recognition apparatus using the same.

The electronic warfare support system is a system for detecting and identifying radar signals in a high density electromagnetic wave signal environment. For the detection, the electronic warfare support system receives the radar signals received from all directions and measures the pulse unit variable specifications in real time, and each radar signal having continuity, regularity and correlation of the signals in the mixed pulse train data mixed with multiple signals. Separate the circles. In addition, the electronic warfare support system analyzes the inter-pulse and intra-pulse modulation characteristics of the pulse train of each radar signal source, and finally compares the radar signal with a built-in identification library.

In order to identify the radar signal, the electronic warfare support system may perform an analysis on a frequency, a pulse repetition interval (PR), a scan, and the like, which are the main factors of the radar signal.

On the other hand, since each radar uses a unique modulation type and value, the PRI is the most basic and essential element used to identify the radar.

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

The method using the histogram is a method of generating a histogram using a difference of time of arrival information (TOA) of each pulse, and recognizing a PRI modulation type by using characteristics of the histogram for each modulation type. The method has a disadvantage in that it is difficult to automate due to the ambiguity of the histogram bin, the threshold, and the like, and the large sensitivity to signal distortion such as missing pulses and unnecessary signals. Thus, the method is mainly used for manual analysis of the operator.

The method using autocorrelation is a method of calculating autocorrelation between pulses and defining a shape separator for distinguishing each PRI modulation type and applying it to each pulse train. The method using the autocorrelation should precede filtering that serves as a preprocessing for the missing and distorted phenomena. In addition, the method using the autocorrelation has a disadvantage in that the time complexity is high because a sufficient number of pulses must be collected to confirm the periodicity since the periodicity is used to distinguish the types, and the autocorrelation is calculated for each pulse.

Accordingly, there is a need to introduce a PRI modulation type recognition method that is robust to signal distortion phenomena such as missing pulses and unnecessary signals and is easy to automate.

Accordingly, the present specification aims at providing measures to solve the above-mentioned problems.

Specifically, an object of the present disclosure is to provide a modulation type recognition method having low sensitivity to missing pulses and distortion pulses.

In addition, the present specification aims at a method for recognizing a modulation type, which is easy to automate due to a simple algorithm, and has a low time complexity.

In addition, an object of the present invention is to provide a modulation mode recognition method with improved modulation mode recognition rate.

In the pulse repetition period modulation type recognition method according to an aspect of the present invention, obtaining an input pulse train, obtaining a chain code for the obtained input pulse train, the state transition probability matrix based on the obtained chain code Acquiring a pulse repetition interval (PRI) modulation type classification factor based on the obtained state transition probability matrix; and obtaining the obtained state transition probability matrix and the obtained PRI modulation type classification factor. Acquiring an output result for a predetermined neural network on the basis of the above, and obtaining a PRI modulation type based on the obtained output result.

Pulse repetition period modulation type recognition method according to another aspect of the present invention may further comprise the step of training the neural network.

In the pulse repetition period modulation type recognition method according to another aspect of the present invention, the step of training the neural network, obtaining a training set, obtaining a chain code for the training pulse train included in the acquired training set Acquiring a state transition probability matrix based on the obtained chain code, acquiring a pulse repetition interval (PRI) modulation type classification factor based on the acquired state transition probability matrix, and The method may include obtaining an output result for a predetermined neural network based on a state transition probability matrix and the obtained PRI modulation type classification factor, and compensating the predetermined neural network based on the obtained output result.

In the pulse repetition period modulation type recognition method according to another aspect of the present invention, the neural network complementing step, the step of determining whether the obtained output result corresponds to the modulation type code of the training pulse train included in the acquired training set It may further include.

In the pulse repetition period modulation type recognition method according to another aspect of the present invention, the neural network complementation step may further include determining whether a residual training set exists.

In a pulse repetition period modulation type recognition method according to another aspect of the present invention, the acquiring of the chain code may include comparing the chain code by comparing a difference between arrival information of the input pulse string and predetermined thresholds for distinguishing the chain code. The method may further include obtaining.

In the pulse repetition period modulation type recognition method according to another aspect of the present invention, obtaining the PRI modulation type, based on the neural network output result, obtaining a result code, based on the obtained result code The method may further include obtaining a PRI modulation form.

A pulse repetition period modulation type recognition apparatus according to an aspect of the present invention includes a chain code obtaining unit obtaining a chain code for an input pulse string, and a state transition probability matrix obtaining unit obtaining a state transition probability matrix based on the chain code. And a PRI modulation type division factor obtaining unit for obtaining a Pulse Repetition Interval (PRI) modulation type division factor based on the state transition probability matrix, and predetermined based on the state transition probability matrix and the PRI modulation type division factor. And a modulation type determination unit configured to obtain an output result of the neural network and determine a PRI modulation type based on the obtained output result.

The pulse repetition period modulation type recognition apparatus according to another aspect of the present invention may further include a neural network training unit for training the neural network.

Pulse repetition period modulation type recognition method according to an aspect of the present invention, obtaining a training set, obtaining a chain code for the training pulse train included in the acquired training set, based on the obtained chain code Obtaining a state transition probability matrix, acquiring a pulse repetition interval (PRI) modulation type classification factor based on the obtained state transition probability matrix, the obtained state transition probability matrix and the obtained PRI Obtaining an output result for a predetermined neural network based on a modulation type classification factor, and comparing the obtained output result with a modulation type code of a training pulse train included in the acquired training set.

The pulse repetition period modulation type recognition method according to another aspect of the present invention may further include supplementing the predetermined neural network based on the comparison result.

Pulse repetition period modulation type recognition method according to an aspect of the present invention, the chain code acquisition unit for obtaining a chain code for the training pulse train included in the training set, obtaining a state transition probability matrix based on the obtained chain code A state transition probability matrix acquisition unit, a PRI modulation type division factor acquisition unit obtaining a pulse repetition interval (PRI) modulation type classification factor based on the obtained state transition probability matrix, the obtained state transition probability matrix and A neural network training unit for acquiring an output result for a predetermined neural network based on the obtained PRI modulation type discrimination factor, and comparing the obtained output result with a modulation type code of a training pulse train included in the acquired training set; Can be.

Disclosure of the present invention solves the problems of the prior art described above.

Specifically, according to the disclosure of the present specification, it is possible to provide a user with a modulation type recognition method having low sensitivity to missing pulses and distortion pulses.

In addition, according to the disclosure of the present specification, it is possible to provide a user with a method for recognizing a modulation type, which algorithm is simple, easy to automate, and low in time complexity.

In addition, according to the disclosure of the present specification, it is possible to provide a user with a modulation mode recognition method having an improved modulation mode recognition rate.

1 is a conceptual diagram illustrating a neural network based PRI modulation type recognition apparatus using a state transition probability according to an embodiment of the present invention.
2 is a flowchart illustrating a PRI modulation type recognition method according to an embodiment of the present invention.
3 is a flowchart illustrating a neural network training method according to an embodiment of the present invention.
4 is a view for explaining a chain code acquisition method according to an embodiment of the present invention.
5 is a diagram for describing a method for obtaining a state transition probability matrix according to an embodiment of the present invention.
6 is a view for explaining a neural network output result obtaining method according to an embodiment of the present invention.
7 is a flowchart illustrating another example of a neural network training method.

It is noted that the technical terms used herein are used only to describe specific embodiments and are not intended to limit the invention. It is also to be understood that the technical terms used herein are to be interpreted in a sense generally understood by a person skilled in the art to which the present invention belongs, Should not be construed to mean, or be interpreted in an excessively reduced sense. In addition, when the technical terms used herein are incorrect technical terms that do not accurately represent the spirit of the present invention, it should be replaced with technical terms that can be understood correctly by those skilled in the art. In addition, the general terms used in the present invention should be interpreted according to a predefined or prior context, and should not be construed as being excessively reduced.

Also, the singular forms "as used herein include plural referents unless the context clearly dictates otherwise. In the present application, the term "comprising" or "comprising" or the like should not be construed as necessarily including the various elements or steps described in the specification, Or may be further comprised of additional components or steps.

In addition, the suffixes "module", "unit", "group", and "unit" for the components used in the present specification are given or used in consideration of ease of specification, meanings distinguished from each other or It does not have a role.

Furthermore, terms including ordinals such as first, second, etc. used in this specification can be used to describe various elements, but the elements should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings, wherein like reference numerals refer to like or similar elements throughout the several views, and redundant description thereof will be omitted.

In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail. It is to be noted that the accompanying drawings are only for the purpose of facilitating understanding of the present invention, and should not be construed as limiting the scope of the present invention with reference to the accompanying drawings.

1

1 is a conceptual diagram illustrating a neural network based PRI modulation type recognition apparatus using a state transition probability according to an embodiment of the present invention.

Referring to FIG. 1, the electronic warfare supporting apparatus S may include a signal receiver 110, a signal processor 120, and a pulse repetition period modulation type recognition apparatus 200.

The signal receiver 110 is configured to receive a radar signal, and the signal processor 120 is configured to process the received radar signal. The signal may be processed in various forms, and for example, the signal processor 120 may be configured to separate the radar signal sources from the received and measured data and extract a pulse train for each signal source. Also, it is possible to identify each signal source by comparing the final analyzed result with the built-in identification information.

Referring to this figure, the pulse repetition period modulation type recognition apparatus 200 is formed to recognize the modulation form of the pulse repetition period from the received radar signal. More specifically, the pulse repetition period modulation type recognition apparatus 200 may include a chain code acquirer 210, a state transition probability matrix acquirer 220, a PRI modulation type classification factor obtainer 230, and a modulation type determiner 240. ), The neural network training unit 250, the neural network processing unit 260, and the like.

The chain code acquisition unit 210 may acquire a chain code for a pulse train. For example, the chain code obtaining unit 210 may generate a chain code for the input pulse string received by the signal receiving unit 110.

The state transition probability matrix acquisition unit 220 may obtain a state transition probability matrix based on a chain code. For example, the state transition probability matrix acquisition unit 220 may generate a state transition probability matrix based on the chain code generated by the chain code acquisition unit 210.

The PRI modulation type classification factor obtaining unit 230 may obtain a PRI modulation type classification factor based on the state transition probability matrix. The PRI modulation type classification factor is an additional factor for PRI modulation type classification.

The neural network processor 260 may obtain a neural network output result based on each element of the state transition probability matrix and a PRI modulation form classification factor.

The modulation type determination unit 240 may determine the PRI modulation type based on the output result of the neural network.

The neural network training unit 250 may include a function for training the neural network.

2

2 is a flowchart illustrating a PRI modulation type recognition method according to an embodiment of the present invention.

According to an embodiment of the present invention, the modulation type recognition apparatus may acquire a predetermined neural network (S210).

Neural networks have been developed over the last half century as a method of computing through emulation of arrays of biological neurons. Thus, in general, neural networks are implemented by processing a combination of many parallel simple calculations.

 Neural networks, on the other hand, can be trained to learn arithmetic functions, algorithm functions, and the like (and thus replicate those functions). At this time, by applying a pattern of inputs to the inputs of the neural network, calculating a difference between desired outputs and outputs at the output nodes of the neural network, and using the difference to modify the weight values of the neural network. Training can be done.

The predetermined neural network is a neural network in which a training process has been performed.

Hereinafter, a process of training a predetermined neural network used in an embodiment of the present invention will be described first.

3

3 is a flowchart illustrating a neural network training method according to an embodiment of the present invention.

According to one embodiment of the invention, the neural network training unit 250 may obtain a training set (S310).

The training set may include a predetermined neural network, a training pulse sequence used to train the predetermined neural network, and a modulation type code for the training pulse sequence. There may be a plurality of training sets.

On the other hand, according to an embodiment of the present invention, the chain code acquisition unit 210 may obtain a chain code for the training set (S320).

4

4 is a view for explaining a chain code acquisition method according to an embodiment of the present invention.

The chain code acquisition unit 210 may acquire a chain code for a pulse train. For example, the chain code obtaining unit 210 may generate a chain code for the input pulse string received by the signal receiving unit 110. In addition, the chain code acquisition unit 210 may generate a chain code for the training pulse train.

The chain code obtaining unit 210 may obtain a difference time of arrival (DTOA) of time of arrival of a pulse train.

In addition, the chain code obtaining unit 210 may include predetermined threshold information. The threshold information is information for generating a chain code based on the DTOA.

For example, referring to FIG. 4, the chain code acquisition unit 210 may include three threshold values TH _ic , TH _dc , and TH _stb . At this time, since the DTOA cannot be negative, only chain codes of +90 degrees to -90 degrees can be generated. And, the generation range of the code is -2 to +2 as shown in FIG. That is, the chain code obtaining unit 210 may generate a chain code (for example, -2, -1, 0, +1, +2) based on the DTOA.

Meanwhile, according to an embodiment of the present invention, the state transition probability matrix acquisition unit 220 may obtain a state transition probability matrix based on the chain code (S330).

5

5 is a diagram for describing a method for obtaining a state transition probability matrix according to an embodiment of the present invention.

The state transition probability matrix acquisition unit 220 may obtain a state transition probability matrix based on a chain code. For example, the state transition probability matrix acquisition unit 220 may generate a state transition probability matrix based on the chain code generated by the chain code acquisition unit 210. In addition, the state transition probability matrix acquisition unit 220 may generate a state transition probability matrix for the training pulse train.

The state transition probability matrix is a matrix that can express the probability of changing from one state (-2, -1, 0, +1, +2) to the next state of the chain code. Referring to FIG. 5, for example, a _{-2, -2} is a probability indicating a probability of changing from -2 state to -2 state, and a _{-1, -2} is a probability indicating a probability of changing from -1 state to -2 state to be.

The state transition probability matrix acquisition unit 220 may read the chain code values one by one and update the matrix to generate the state transition probability matrix.

Meanwhile, according to an embodiment of the present invention, the PRI modulation type classification factor obtaining unit 230 may obtain a PRI modulation type classification factor based on the state transition probability matrix (S340).

The PRI modulation type classification factor is an additional factor for PRI modulation type classification. For example, the PRI modulation pattern classification factor may include a contrast factor (P _cont ), a homogeneity factor (P _homo ), and a momentum factor (P _momt ).

The contrast factor P _cont is a factor that can indicate a difference in state change and can be extracted by the following equation.

The homogeneity factor (P _homo ) is a factor that can represent the similarity of state changes, and can be extracted by the following equation.

The momentum factor P _momt is a factor that may indicate monotony of a state change, and may be extracted by the following equation.

Meanwhile, according to an embodiment of the present invention, the neural network processing unit 260 may obtain a neural network output result based on each element of the state transition probability matrix and a PRI modulation form classification factor (S350).

6

6 is a view for explaining a neural network output result obtaining method according to an embodiment of the present invention.

Referring to FIG. 6, the neural network processor 260 may input values of 25 elements and 3 PRI modulation type classification factors of the state transition probability matrix.

According to an embodiment of the present invention, as shown in FIG. 6, a three-layer Multi Layer Perceptron (MLP) having 28 input layers, eight hidden layers, and seven output layers may be used. In addition, a binary sigmoid function (τ) can be used as an activation function, and its equation and differential property are as follows.

As described above, the neural network processor 260 may obtain an output for input (25 elements of the state transition probability matrix and 3 PRI modulation type classification factors) by using the neural network illustrated in FIG. 6.

Meanwhile, according to an embodiment of the present invention, the neural network training unit 250 may complement the neural network based on the neural network output result (S360).

Compensating the neural network may include determining whether a neural network output result corresponds to a modulation type code of a training pulse train included in a training set.

In addition, the supplementing of the neural network may include determining whether a residual training set exists.

A detailed complementary method will be described in detail with reference to FIG. 7.

7

7 is a flowchart illustrating another example of a neural network training method.

Referring to FIG. 7, the neural network training unit 250 may obtain a training set (S710).

In addition, according to an embodiment of the present invention, the chain code acquisition unit 210 may acquire a chain code for an arbitrary training set (S720). In this case, since there may be a plurality of training sets, the following process may be performed on some of the plurality of training sets.

In addition, according to an embodiment of the present invention, the state transition probability matrix acquisition unit 220 may obtain a state transition probability matrix based on the obtained chain code (S730).

In addition, according to an embodiment of the present invention, the PRI modulation type classification factor acquisition unit 230 may obtain a PRI modulation type classification factor based on the obtained state transition probability matrix (S740).

In addition, according to an embodiment of the present invention, the neural network processing unit 260 may obtain a neural network output result based on each element of the obtained state transition probability matrix and the obtained PRI modulation type classification factor. It may be (S750).

Since the steps described above with reference to FIG. 3 may be applied, a detailed description thereof will be omitted.

According to an embodiment of the present invention, the neural network training unit 250 may determine whether the neural network output result corresponds to the modulation type code for the training pulse train (S760). For example, the modulation form determination unit 240 may obtain a modulation form code based on the neural network output result, and the neural network training unit 250 may compare the obtained modulation form code with the modulation form code for the training pulse train. .

On the other hand, the specific operation of the modulation type determination unit 240 will be described later.

When the neural network output result and the modulation type code for the training pulse train do not correspond, the neural network training unit 250 may perform weight update (S770).

The weight update checks whether the modulation type code corresponds to the neural network output value and the training pulse train in each training step, and if there is a mismatch, the internal input layer of the neural network using error propagation to derive the modulation type code value. Means updating the weight u between the hidden layer and the hidden layer and the weight v between the hidden layer and the output layer.

On the other hand, according to an embodiment of the present invention, the neural network training unit 250 may determine whether there is a residual training set (S780). Since there may be a plurality of training sets, when the training process is performed for some training sets, the above-described series of processes may be performed again for training sets other than the used training sets.

Through the series of processes described with reference to FIGS. 3 to 7, the modulation type recognition apparatus 200 may obtain a neural network to be used for modulation type recognition (S210 and FIG. 2).

According to an embodiment of the present invention, the chain code obtaining unit 210 may obtain an input pulse string and may obtain a chain code for the input pulse string (S220).

In addition, according to an embodiment of the present invention, the state transition probability matrix acquisition unit 220 may obtain a state transition probability matrix based on the obtained chain code (S230).

In addition, according to an embodiment of the present invention, the PRI modulation type classification factor obtaining unit 230 may obtain a PRI modulation type classification factor based on the obtained state transition probability matrix (S240).

Further, according to an embodiment of the present invention, the neural network processing unit 260 is a neural network based on the acquired neural unit, each element of the obtained state transition probability matrix, and the obtained PRI modulation type classification factor. In operation S250, an output result may be obtained.

Except that the above steps are performed on the input pulse train, the contents described above with reference to FIGS. 3 to 6 may be applied, and thus a detailed description thereof will be omitted.

Meanwhile, according to an embodiment of the present invention, the modulation type determination unit 240 may obtain a PRI modulation type based on the neural network output result (S260).

For example, the modulation type determination unit 240 may obtain PRI modulation type information in response to the neural network output result.

Alternatively, the modulation type determination unit may obtain a result code. For example, the modulation type determination unit 240 may find an index having the maximum value among the neural network output result values and generate a result code based on the index. For example, referring to FIG. 6, the output values of the neural network are all seven. At this time, if the third value of the result value has a maximum value, the result code may be generated as [0010000], and if the fourth value of the result value has the maximum value, the result code may be generated as [0001000]. .

The modulation type determination unit 240 may recognize the PRI modulation type based on the result code. For example, the modulation type determination unit 240 may recognize the PRI modulation type by comparing the predetermined PRI modulation code with the result code. The PRI modulation code may be defined as shown in Table 1.

PRI modulation mode

Modulation code
Dwell & Switch

1000000
Linear sliding (+)

0100000
Linear sliding (-)

0010000
Nonlinear Sliding (+)

0001000
Non-linear sliding (-)

0000100
Wobble

0000010
Jitter

0000001

Through this process, the PRI modulation mode recognition apparatus of the present invention can recognize the PRI modulation type for the input pulse train.

The present invention is a modulation type recognition method having low sensitivity to missing and distortion pulses by generating state transition probability and additional PRI modulation type discrimination factors. Therefore, the algorithm is simple, easy to automate, low in time complexity, and high in modulation type recognition rate, so that it can be directly applied to an electronic warfare support (ES) system and an electronic intelligence (ELINT) system requiring signal analysis.

The above-described method according to the embodiment of the present invention can be used individually or in combination. Further, the steps constituting each embodiment can be used individually or in combination with the steps constituting the other embodiments.

In addition, the above-described method can be implemented in a recording medium readable by a computer or the like using, for example, software, hardware, or a combination thereof.

According to a hardware implementation, the methods described so far can be applied to various types of application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays processors, controllers, micro-controllers, microprocessors, and other electronic units for performing other functions.

According to a software implementation, the procedures and functions described herein may be implemented in separate software modules. The software modules may be implemented in software code written in a suitable programming language. The software code may be stored in a storage unit and executed by a processor.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It falls within the scope of the invention.

200: modulation type recognition device
210: chain code acquisition unit
220: state transition probability matrix acquisition unit
230: PRI modulation type classification factor acquisition unit
240: modulation type determination unit
250: neural network training unit
260: neural network processing unit

Claims (9)

Obtaining an input pulse train;
Obtaining a chain code for the obtained input pulse train;
Obtaining a state transition probability matrix based on the obtained chain code;
Acquiring a Pulse Repetition Interval (PRI) modulation type classification factor based on the obtained state transition probability matrix;
Obtaining an output result for a predetermined neural network based on the obtained state transition probability matrix and the obtained PRI modulation type classification factor; And
And acquiring a PRI modulation form based on the obtained output result. The method of claim 1,
Further comprising training the neural network,
Training the neural network,
Obtaining a training set;
Acquiring a chain code for a training pulse train included in the acquired training set;
Obtaining a state transition probability matrix based on the obtained chain code;
Acquiring a Pulse Repetition Interval (PRI) modulation type classification factor based on the obtained state transition probability matrix;
Obtaining an output result for a predetermined neural network based on the obtained state transition probability matrix and the obtained PRI modulation type classification factor; And
And replenishing the predetermined neural network based on the obtained output result. The method of claim 2, wherein the neural network complementary step,
And determining whether the obtained output result corresponds to a modulation type code of a training pulse train included in the acquired training set. The method of claim 2, wherein the neural network complementary step,
And determining whether there is a residual training set. The method of claim 1, wherein the obtaining of the chain code comprises:
And comparing the difference between the arrival information of the input pulse string and predetermined thresholds for distinguishing the chain code, to obtain the chain code. The method of claim 1, wherein the obtaining of the PRI modulation form comprises:
Obtaining a result code based on the neural network output result; And
And acquiring a PRI modulation type based on the obtained result code. A chain code obtaining unit obtaining a chain code for an input pulse train;
A state transition probability matrix obtaining unit obtaining a state transition probability matrix based on the chain code;
A PRI modulation type classification factor obtaining unit obtaining a pulse repetition interval (PRI) modulation type classification factor based on the state transition probability matrix; And
And a modulation type determination unit for obtaining an output result for a predetermined neural network based on the state transition probability matrix and the PRI modulation type classification factor, and determining a PRI modulation type based on the obtained output result. Pulse repetition period modulation type recognition device. Obtaining a training set;
Acquiring a chain code for a training pulse train included in the acquired training set;
Obtaining a state transition probability matrix based on the obtained chain code;
Acquiring a Pulse Repetition Interval (PRI) modulation type classification factor based on the obtained state transition probability matrix;
Obtaining an output result for a predetermined neural network based on the obtained state transition probability matrix and the obtained PRI modulation type classification factor; And
And comparing a modulation form code of a training pulse train included in the obtained training set with the obtained output result. A chain code obtaining unit obtaining a chain code for a training pulse train included in a training set;
A state transition probability matrix obtaining unit obtaining a state transition probability matrix based on the obtained chain code;
A PRI modulation type classification factor obtaining unit obtaining a pulse repetition interval (PRI) modulation type classification factor based on the obtained state transition probability matrix; And
Acquiring an output result for a predetermined neural network based on the obtained state transition probability matrix and the obtained PRI modulation type classification factor, and a modulation type code of a training pulse train included in the acquired output result and the acquired training set Pulse repetition period modulation type recognition method comprising a neural network training unit for comparing the.

Images