KR101910540B1 - Apparatus and method for recognizing radar waveform using time-frequency analysis and neural network - Google Patents

Abstract

According to an embodiment of the present invention, in radar modulation form recognition using time-frequency analysis and a neural network, a sampling decrease rate of a detected target radar signal is set. A frequency energy distribution image with respect to time is obtained by using CWD among time-frequency analysis methods for the detected target radar signal. The obtained image is inputted into a convolution neural network to recognize a modulation form of the target radar signal to automatically, quickly, and accurately recognize modulation forms of modulation and continuous wave radar signals within pulses of 12 modulation methods including linear frequency modulation (LFM), Costas code, binary phase shift keying (BPSK), polyphase (Frank, P1, P2, P3, P4), and polytime (T1, T2, T3, T4) even in a high noise environment and easily recognize modulation forms of radar signals of other modulation methods.

Description

TECHNICAL FIELD [0001] The present invention relates to a radar modulation type recognition apparatus and method using time frequency analysis and a neural network,

The present invention relates to an electronic warfare system, and more particularly, to an apparatus and method for recognizing a modulated form of a radar signal using time frequency analysis and a neural network in an electronic warfare support system.

A system for detecting and identifying radar signals in electronic warfare support systems is one of the important technologies.

In order to accurately detect and identify the radar signal in the above system, after receiving each radar signal source, it is necessary to analyze the characteristics of the corresponding radar signal from the received radar signal and obtain the identification factor.

At this time, the modulated form of each radar signal source is one of important identification parameters for identification of the radar signal, and identification factors such as the period, frequency, and the like of the radar signal can be obtained based on the obtained radar modulation form.

Finally, the radar signals are identified by comparing the identification factors obtained as described above with the built-in identification library stored in advance.

However, in recent years, due to the gradual development of radar technology and the rapid increase of all equipment using electromagnetic signals, it becomes more difficult to accurately identify radar signal sources.

Conventionally, as a method of recognizing a modulated form of a radar signal for identification of a radar signal as described above, various characteristic factors that can be obtained from a radar signal, such as a Gini coefficient value or an instantaneous frequency of a radar signal power spectrum and a statistical test, A method of recognizing a modulated form of a radar signal has been proposed.

However, in the method of recognizing a modulated form of a radar signal using conventional feature parameters, only a specific modulated form of a radar signal can be identified. In order to identify a radar signal that is not identified by known feature factors, The process of finding the factor is necessary.

In addition, since the receiving environment of the radar signal is generally a noisy environment, conventional characteristic factors are not accurately extracted in the environment with a lot of signal noises, and the identification performance of the radar signal is deteriorated.

Korean Registered Patent No. 10-1235059 (Registered Date Feb. 13, 2013)

Accordingly, in one embodiment of the present invention, a frequency energy distribution image is obtained with respect to the target radar signal detected in the electronic warfare support system using a time frequency analysis method, and the obtained image is divided into a series of convolution neural networks And to provide a radar modulation shape recognition apparatus and method using a time frequency analysis and a neural network to recognize a modulated form of a target radar signal.

A radar signal receiving unit for converting a target radar signal into a digital signal and generating a discrete radar signal; and a radar signal receiving unit for converting the preprocessed discrete radar signal An image preprocessing unit for performing preprocessing to reduce the capacity of the time-frequency image; and an image processing unit for processing the preprocessed time-frequency image with the time-frequency image and the time- And a modulation type recognition unit for inputting the input signal to a classifier of a Convolution Neural Network (CNN) system that has studied the relationship between the shapes and analyzing the modulation type of the target radar signal.

In addition, the time-frequency analysis processing unit may generate the time-frequency image using the time-frequency analysis technique as the time-frequency analysis algorithm.

The image preprocessing unit reconstructs the image by replacing a brightness value of each pixel on the time-frequency image with a brightness value of each pixel by dividing a brightness value of each pixel on the time-frequency image by a maximum value of brightness values of the pixels on the time- .

The modulation type recognizing unit is configured to learn the convolutional neural network by inputting training time frequency images for a plurality of learning radar signals having different modulation types into the convolutional neural network.

In addition, the time-frequency image may be an energy distribution of the frequency-dependent frequency of the preprocessed discrete radar signal as an image.

According to another aspect of the present invention, there is provided an apparatus for recognizing a radar modulation type, comprising: a radar signal receiving unit for converting a target radar signal into a digital signal to generate a discrete radar signal; A sampling determination unit for determining whether sampling is performed on the abnormal radar signal and a sampling reduction ratio at the time of executing the sampling; and a preprocessing unit for performing preprocessing on the discrete radar signal only when the sampling is performed, A time-frequency analysis processor for generating a time-frequency image of the preprocessed discrete radar signal using a time-frequency analysis algorithm; Pre-processing to reduce capacity And inputting the preprocessed time-frequency image to a classifier of the Convolution Neural Network (CNN) series which learns the relationship between the time-frequency image and the modulation form through a learning time-frequency image, And a modulation type recognizing unit for analyzing the modulated form of the radar signal.

The sampling determination unit may calculate the sampling reduction ratio indicating whether the number of samples of the target radar signal is reduced to several times using the maximum frequency value, and if the sampling reduction ratio is calculated to be equal to or greater than a predetermined ratio And determines not to perform the sampling on the target radar signal when the sampling reduction ratio is less than a predetermined ratio.

The radar signal preprocessing unit may calculate an average value of a plurality of consecutive samples in the discrete radar signal and replace the plurality of samples with one sample having the calculated average value to calculate a number of samples of the discrete radar signal And decreasing the sampling rate according to the sampling reduction ratio.

In addition, the time-frequency analysis processing unit may generate the time-frequency image using the time-frequency analysis technique as the time-frequency analysis algorithm.

Also, the time-frequency image is characterized by representing an energy distribution by frequency of the pre-processed discrete radar signal over time as an image.

The image preprocessing unit reconstructs the image by replacing a brightness value of each pixel on the time-frequency image with a brightness value of each pixel by dividing a brightness value of each pixel on the time-frequency image by a maximum value of brightness values of the pixels on the time- .

The modulation type recognizing unit is configured to learn the convolutional neural network by inputting training time frequency images for a plurality of learning radar signals having different modulation types into the convolutional neural network.

When the modulated form of the target radar signal is recognized as a polyphase modulated continuous wave radar signal through the modulation type recognizing unit, the apparatus estimates the center frequency and the period of the target radar signal, And a modulation type recognition auxiliary unit for extracting a code length and a characteristic factor of the target radar signal and recognizing a detailed modulation type of the target radar signal using the code length and the characteristic factor.

The feature parameter is information on the frequency or bandwidth of the radar signal.

In another aspect of the present invention, there is provided a method of recognizing a radar modulation type, comprising: generating a discrete radar signal by converting a target radar signal into a digital signal when the target radar signal is received; Determining whether to perform sampling on the abnormal radar signal and a sampling reduction ratio at the time of executing the sampling; and decreasing the sampling number of the discrete radar signal to the predetermined sampling reduction ratio when the sampling is determined to be performed Generating a time-frequency image for the discrete radar signal using a time-frequency analysis algorithm; performing pre-processing to reduce the capacity of the time-frequency image; The time-frequency image and the The input to the convolutional neural network learning the relationship between the form and a step of recognizing the modulation type of the target radar signal.

The determining step may further include calculating the sampling reduction ratio indicating whether the number of samples of the target radar signal is reduced to several times using the maximum frequency value, Determining if the sampling reduction rate is less than a predetermined rate; and determining not to perform sampling on the target radar signal if the sampling reduction ratio is less than a predetermined rate.

The step of decreasing the number of samples of the radar signal may further include calculating an average value of a plurality of consecutive samples in the discrete radar signal and replacing the plurality of samples with one sample having the calculated average value And decreasing the number of samples of the discrete radar signal according to the sampling reduction ratio.

In addition, the time-frequency image is generated using the time-frequency analysis technique as the time-frequency analysis algorithm.

The step of performing image preprocessing may include replacing a brightness value of each pixel on the time-frequency image with a brightness value of each pixel by dividing a brightness value of each pixel on the time-frequency image by a maximum brightness value of each pixel, Is reconstructed.

The recognizing may include learning the convolutional neural network by inputting a training time frequency image for a plurality of learning radar signals having different modulation types into a convolutional neural network, And recognizing the modulated form of the target radar signal by inputting the modulated signal to the learned convolution neural network.

When the modulated form of the target radar signal is recognized as a polyphase modulated continuous wave radar signal, the method estimates a center frequency and a period of the target radar signal, And recognizing a detailed modulation type of the target radar signal using the code length and the characteristic factor.

According to an embodiment of the present invention, in a radar modulation shape recognition using a time frequency analysis and a neural network, a time frequency analysis method is used for a detected target radar signal to obtain a frequency energy distribution image over time, (Costar Code), Binary Phase Shift Keying (BPSK), Polyphase (Frank, P1, P2, P3, P4), and Polytime (T1, T2, T3, T4), and the modulation type of the continuous wave radar signal can be automatically recognized. At the same time, the modulation type of the radar signal of another modulation method can be easily recognized.

According to an embodiment of the present invention, by reducing the sampling number of the discrete radar signal to a predetermined sampling reduction rate and using the time frequency analysis and the CNN class classifier, the radar modulation form can be quickly Can be accurately recognized.

1 is a functional block diagram of a radar modulation shape recognition apparatus using a time frequency analysis and a neural network according to an embodiment of the present invention.
2 is a flowchart illustrating an operation control of a radar signal preprocessing unit according to an embodiment of the present invention;
3 is an illustration of a time-frequency image corresponding to each of the modulated forms of modulated radar signals in the 12 pulses analyzed via CWD in accordance with an embodiment of the present invention.
4 is an illustration of a time-frequency image corresponding to each of the modulated forms of 12 sustained wave radar signals analyzed via CWD in accordance with one embodiment of the present invention.
5 is a flowchart showing an operation control of an image preprocessing unit according to an embodiment of the present invention.
6 is a flowchart of operation control of a modulation type recognizing unit according to an embodiment of the present invention;

Hereinafter, the operation principle of the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. The following terms are defined in consideration of the functions of the present invention, and these may be changed according to the intention of the user, the operator, or the like. Therefore, the definition should be based on the contents throughout this specification.

1 is a functional block diagram of a radar modulation shape recognition apparatus using a time frequency analysis and a neural network according to an embodiment of the present invention.

Referring to FIG. 1, a radar modulation pattern recognition apparatus 100 using a time frequency analysis and a neural network according to an embodiment of the present invention includes a radar signal reception unit 102, a sampling determination unit 104, a radar signal preprocessing unit 106, A temporal frequency analysis processing unit 108, an image preprocessing unit 110, a modulation type recognition unit 112, a modulation type recognition assistance unit 114, and the like.

First, when receiving a target radar signal requiring modulation type recognition, the radar signal receiving unit 102 down-converts the frequency of the received target radar signal and then samples the down-converted analog target radar signal. And converts it into a digital signal. Accordingly, the target radar signal can be converted into a discrete radar signal through digital conversion.

At this time, the discrete radar signal y [k] converted into the digital signal through the radar signal receiving unit 102 can be expressed as Equation (1) below.

는 수신된 복조 이산 레이더 신호이고,

는 부가 백색 가우스 잡음(AWGN, additive white gaussian noise)이고,

는 0이 아닌 상수를 의미할 수 있다. 또한,

는 샘플링 비율(sampling rate)이며,

는

마다 얻어지는 샘플링 인덱스(sample index)를 나타낼 수 있다. here,

Is a received demodulated discrete radar signal,

Is additive white gaussian noise (AWGN)

May refer to a non-zero constant. Also,

Is the sampling rate,

The

A sample index can be obtained.

는 타겟 레이더 신호의 순간 위상(instantaneous phase)을 나타내는 것으로 아래의 [수학식 2]와 같이 나타낼 수 있다.Also,

Represents the instantaneous phase of the target radar signal, and can be expressed by Equation (2) below.

는 순간 주파수(instantaneous frequency)이고,

는 위상 변이(phase offset)이다.In the above equation (2)

Is an instantaneous frequency,

Is the phase offset.

Next, the sampling determination unit 104 selects a predetermined plurality of consecutive samples for the discrete data signal generated from the radar signal receiving unit 102, determines whether or not to use a technology for averaging the sizes of the selected plurality of samples And sets the sampling reduction ratio at the same time.

Next, the radar signal preprocessing unit 106 determines whether the condition for performing the preprocessing on the discrete radar signal generated from the radar signal receiving unit 102 is satisfied or not, from the sampling determination unit 104, Reduces the number of radar signals to a predetermined sampling reduction rate.

2 shows a flow of operation control of the radar signal preprocessing unit 106 according to an embodiment of the present invention. Hereinafter, operations of the sampling determination unit 104 and the radar signal preprocessing unit 106 will be described in more detail with reference to FIG.

The sampling determination unit 104 receives the digitally processed discrete radar signal y [k] from the radar signal reception unit 102 (S200).

Then, the sampling determining unit 104 determines whether to reduce the number of samples by estimating the maximum frequency value of the received target radar signal, and determines a sampling reduction ratio when the number is reduced.

)는 다음과 같은 [수학식 3]으로 나타낼 수 있다. At this time, the estimated maximum frequency (

) Can be expressed by the following equation (3).

는 수신된 레이더 신호의 순간 주파수의 최대값이고,

는 중간 주파수이고,

는 레이더 신호 수신부(102)의 대역필터(band pass filter)의 대역폭(bandwidth)이다.In the above equation (3)

Is the maximum value of the instantaneous frequency of the received radar signal,

Is an intermediate frequency,

Is a bandwidth of a band pass filter of the radar signal receiving unit 102.

Further, the sampling reduction ratio can be obtained using Equation (4) by using the maximum value of the received radar signal instantaneous frequency obtained in Equation (3).

은 한 개의 반송파(carrier wave)를 나타내는 샘플(sample)의 개수를 의미하고,

는 한 개의 반송파를 나타내는 데 필요한 최소 샘플의 개수를 의미하고,

는 샘플의 개수를 몇 배로 줄일지를 나타낸다.In Equation (4) above,

Denotes the number of samples representing one carrier wave,

Denotes the minimum number of samples required to represent one carrier,

Indicates how many times the number of samples is reduced.

값을 추정하고, 미리 기설정된

와

으로부터 샘플링 감소 비율인

를 구할 수 있다. 또한

면 샘플 평균화 기술을 사용하지 않는다.At this time,

Estimates a value,

Wow

Lt; RTI ID = 0.0 >

Can be obtained. Also

No sample averaging technique is used.

라는 조건을 만족하면, 샘플링 결정부(104)에서 샘플링 실행 여부를 결정하게 되고, 이에 따라 레이더 신호 전처리부(106)는 평균값을 이용하여 이산 레이더 신호의 샘플링 개수를 줄인다(S204). In other words,

The sampling determination unit 104 determines whether sampling is to be executed. Accordingly, the radar signal preprocessing unit 106 reduces the sampling number of the discrete radar signal using the average value (S204).

For example, in the operation of reducing the number of sampling of the discrete radar signal, the radar signal preprocessing unit 106 selects a predetermined plurality of consecutive samples for the digitally converted discrete radar signal, And calculates an average value for each of them. Subsequently, one sample having an average value calculated for a plurality of samples is substituted. Accordingly, for example, when four samples are selected as a single sample, the number of samples of the discrete radar signal is reduced to 1/4, and the subsequent time frequency analysis processing unit 108 calculates a frequency The calculation process for analyzing the energy distribution can be further simplified.

A method of selecting a predetermined number of discrete radar signals as described above and calculating an average value of the magnitudes of the selected signals to generate a single signal is performed by using a time frequency analysis noise noise portion of the signal has a relatively larger value corresponding to the signal, so that the noise can be stronger.

) 만큼 샘플 개수가 줄어든 전처리된 이산 레이더 신호(

)는 아래의 [수학식 5]와 같이 나타낼 수 있다.At this time, the ratio set by the method using the average value of the discrete radar signals

), A preprocessed discrete radar signal (

) Can be expressed by the following equation (5).

As described above, the preprocessed discrete radar signal may be applied to the time-frequency analysis processing unit 108 and used to analyze the frequency energy distribution according to the time of the received radar signal.

Next, the time-frequency analysis processing unit 108 generates a time-frequency image using a time-frequency analysis algorithm for the discrete radar signal pre-processed by the radar signal preprocessing unit 106. This time-frequency image is an image containing information on the energy distribution of the frequency over time of the discrete radar signal.

In this case, the time-frequency analysis processor 108 uses a time-frequency analysis algorithm to analyze the amount of frequency energy over time. Here, the Choi-William distribution (CWD) algorithm But the present invention is not limited thereto.

)는 아래의 [수학식 6]과 같이 계산되어 생성될 수 있다.Here, the time-frequency image for the target radar signal via CWD (

Can be calculated and generated as shown in Equation (6) below.

과

는 시간과 주파수 인덱스이고,

은 샘플의 수이고,

와

는 크기가 일정한 윈도우이다.In the above equation (6)

and

Is a time and frequency index,

Is the number of samples,

Wow

Is a window of constant size.

3 is an illustration of a time-frequency image corresponding to each of the modulated forms of modulated radar signals in the 12 pulses analyzed via CWD in accordance with one embodiment of the present invention.

4 is an illustration of a time-frequency image corresponding to each of the modulated forms of 12 sustained wave radar signals analyzed via CWD in accordance with an embodiment of the present invention.

Referring to FIGS. 3 and 4, the modulation patterns of the twelve pulse modulated and continuous wave radar signals include, for example, LFM (Linear Frequency Modulation), Costas Code, Binary Phase Shift Keying (BPSK), Polyphase , P2, P3, P4), Polytime (T1, T2, T3, T4), and the like. Meanwhile, in FIGS. 3 and 4, the modulation patterns of the 12 pulses and the continuous wave radar signals are illustrated. However, it is needless to say that 12 or more modulation patterns can be analyzed.

Next, the image preprocessing unit 110 preprocesses the time-frequency image with respect to the target radar signal generated by the time-frequency analysis processing unit 108.

5 is a flowchart illustrating an operation control of the image preprocessing unit 110 according to an embodiment of the present invention. Hereinafter, the operation of the image preprocessing unit 110 will be described in more detail with reference to FIG.

)를 입력받는다(S500).The image preprocessing unit 110 prepares a time-frequency image for the target radar signal generated from the time-frequency analysis processing unit 108

(S500).

Next, a normalization process is performed in which a maximum value among the brightness values of each pixel is found on the time-frequency image and a brightness value of each pixel is divided by a maximum value into brightness values of each pixel to perform a normalization process, (S504).

Next, the image preprocessing unit 110 reduces the reconstructed time-frequency image to an image size that can be processed by the CNN classifier of the modulation type recognizing unit 112 (S506). At this time, in reducing the image size, the image preprocessing unit 110 determines the degree of reduction of the size of the image in consideration of the degree of loss of information possessed by the time-frequency image and the processing speed in the modulation type recognizing unit 112 But is not limited thereto.

Next, the modulation type recognizing unit 112 recognizes a pre-processed image generated from the image preprocessing unit 110 by using a Convolution Neural Network (hereinafter referred to as " Convolution Neural Network ") in which a relation between a time- CNN) to recognize the modulated form of the target radar signal. On the other hand, the modality type classifier is not limited to CNN but may be classified as a CNN-based neural network.

6 shows a flow of operation control of the modulation type recognizing unit 112 according to an embodiment of the present invention. Hereinafter, the operation of the modulation type recognizing unit 112 will be described in more detail with reference to FIG.

The modulation type recognizing unit 112 reads a time-frequency image of the target radar signal preprocessed by the image preprocessing unit 110 (S600).

Then, the modulation type recognizing unit 112 inputs the time-frequency image of the read-in target radar signal to the CNN classifier, which is already learned, and recognizes the modulation type of the target radar signal (S602).

In this case, the CNN classifier moves the image filter of a predetermined size, extracts features by performing a convolution operation with pre-stored time-frequency images, and classifies the desired results through the extracted features can do.

In the embodiment of the present invention, the modulation type recognizing unit 112 inputs and learns a learning time-frequency image of a plurality of learning radar signals having different modulation types previously known to the CNN classifier, When a time-frequency image is input, the CNN classifier is used to recognize a modulation type of the target radar signal.

When the radar signals are classified into ambiguous radar signals, which are recognized only by a part of time-frequency images, the modulation shape recognition sub-unit 114 extracts feature parameters extracted using the time-frequency image and the radar signal modulation recognition result as a decision tree or a neural network It is input to the classifier and recognized in more detail (S604).

That is, when the result recognized by the modulation type recognizing unit 112 is classified into a polyphase-modulated continuous wave radar signal, the modulation shape recognition sub-unit 114 outputs a signal The characteristics of the radar signal, such as the center frequency, period, bandwidth, and code length of the radar signal, are obtained in detail.

At this time, if the code length of a signal classified as Frank or P1 is odd, it is not necessary to distinguish two continuous wave radar signals because they are the same signal. If the code length of the signal is an even number, it generates the Frank and P1 signals using the feature parameters such as the extracted frequency and the code length, and classifies the received signal into a signal having a high correlation result with the received signal.

Also, if the code length of a signal classified as P3 or P4 is an even number, it is not necessary to distinguish between two continuous wave radar signals because they are the same signal. When the code length of the signal is odd, in the case of P3, only the time-frequency image can be recognized as a characteristic.

As described above, according to the embodiment of the present invention, in the radar modulation type recognition using the time frequency analysis and the neural network, the sampling reduction ratio of the detected target radar signal is set, Among the analysis methods, CWD is used to obtain a frequency energy distribution image over time, and the obtained image is input to a convolutional neural network to recognize the modulation type of the target radar signal. Thus, LFM (Costar Code), BPSK Binary Phase Shift Keying), Polyphase (Frank, P1, P2, P3, P4) and Polytime (T1, T2, T3 and T4) And at the same time, the modulated form of the radar signal of another modulation method can be easily recognized.

Combinations of the steps of each flowchart attached to the present invention may be performed by computer program instructions. These computer program instructions may be loaded into a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus so that the instructions, which are executed via a processor of a computer or other programmable data processing apparatus, Lt; / RTI > These computer program instructions may also be stored in a computer usable or computer readable memory capable of directing a computer or other programmable data processing apparatus to implement the functionality in a particular manner so that the computer usable or computer readable memory It is also possible to produce manufacturing items that contain instruction means for performing the functions described in each step of the flowchart. Computer program instructions may also be stored on a computer or other programmable data processing equipment so that a series of operating steps may be performed on a computer or other programmable data processing equipment to create a computer- It is also possible for the instructions to perform the processing equipment to provide steps for executing the functions described in each step of the flowchart.

In addition, each step may represent a module, segment, or portion of code that includes one or more executable instructions for executing the specified logical function (s). It should also be noted that in some alternative embodiments, the functions mentioned in the steps may occur out of order. For example, the two steps shown in succession may in fact be performed substantially concurrently, or the steps may sometimes be performed in reverse order according to the corresponding function.

While the invention has been shown and described with reference to certain preferred embodiments thereof, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention. Accordingly, the scope of the invention should not be limited by the described embodiments but should be defined by the appended claims.

102: Radar signal reception unit 104: Sampling determination unit
106: Radar signal preprocessing unit 108: Time frequency analysis unit
110: image preprocessing unit 112: modulation type recognizing unit
114: modulation shape recognition assistant

Claims (16)

A radar signal receiving unit for converting the target radar signal into a digital signal to generate a discrete radar signal,
A sampling determination unit for determining whether to perform sampling on the discrete radar signal using a maximum frequency value of the target radar signal and a sampling reduction ratio at the time of executing the sampling;
A radar signal preprocessing unit for performing preprocessing on the discrete radar signal to reduce the number of samples of the discrete radar signal according to the predetermined sampling reduction ratio only when the sampling execution is determined;
A time-frequency analysis processor for generating a time-frequency image of the preprocessed discrete radar signal using a time-frequency analysis algorithm;
An image preprocessing unit for performing preprocessing for reducing the capacity of the time-frequency image;
The pre-processed time-frequency image is input to a classifier of the Convolution Neural Network (CNN) series which learns the relationship between the time-frequency image and the modulation form through the learning time-frequency image, A modulation type recognizing unit for analyzing,
A characteristic factor extracted from the target radar signal when the modulated form of the target radar signal is recognized as a polyphase modulated continuous wave radar signal through the modulation type recognizer, And a modulation type recognition auxiliary unit for recognizing a detailed modulation type of the target radar signal using a code length of the target radar signal
Radar modulation shape recognition device.
The method according to claim 1,
Wherein the sampling determination unit comprises:
Calculating a sampling reduction ratio indicating how many times the number of samples of the target radar signal is reduced by using the maximum frequency value; and sampling the target radar signal when the sampling reduction ratio is not less than a predetermined ratio, And determines not to perform the sampling on the target radar signal if the sampling reduction rate is less than a predetermined rate
Radar modulation shape recognition device.
The method according to claim 1,
The radar signal pre-
Calculating a mean value of a plurality of consecutive samples in the discrete radar signal and replacing the plurality of samples with one sample having the calculated mean value to reduce the number of samples of the discrete radar signal according to the sampling decrease rate
Radar modulation shape recognition device.
The method according to claim 1,
Wherein the time-
Wherein the time-frequency analysis algorithm comprises: generating a time-frequency image using a time-frequency analysis technique
Radar modulation shape recognition device.
The method according to claim 1,
The time-
The energy distribution of the pre-processed discrete radar signal with respect to time is shown by an image.
Radar modulation shape recognition device.
The method according to claim 1,
The image pre-
Wherein the image is reconstructed by replacing a brightness value of each pixel on the time-frequency image with a brightness value of each pixel by dividing a brightness value of each pixel by a maximum value of brightness values of the pixels on the time-
Radar modulation shape recognition device.
The method according to claim 1,
The modulation type recognizing unit recognizes,
A learning time-frequency image for a plurality of learning radar signals having different modulation types is input to the convolutional neural network to learn the convolutional neural network
Radar modulation shape recognition device.
delete delete Generating a discrete radar signal by converting the target radar signal into a digital signal when the target radar signal is received,
Determining whether to perform sampling on the discrete radar signal using a maximum frequency value of the target radar signal and a sampling reduction ratio at the time of executing the sampling;
Decreasing the sampling number of the discrete radar signal to a predetermined sampling reduction rate when the sampling execution is determined;
Generating a time-frequency image for the discrete radar signal using a time-frequency analysis algorithm;
Performing a preprocessing to reduce the capacity of the time-frequency image;
Recognizing a modulation type of the target radar signal by inputting the pre-processed time-frequency image to a convolutional neural network that learns a relationship between the time-frequency image and a modulation form through a learning time-frequency image;
Wherein the characteristic factor includes a center frequency, a period, a bandwidth, and a code length of the target radar signal when the target radar signal is recognized as a polyphase-modulated continuous wave radar signal. - extracting a box,
And recognizing the sub-modulated form of the target radar signal using the feature factor
Radar modulation type recognition method.
11. The method of claim 10,
Wherein the determining comprises:
Calculating the sampling reduction ratio indicating whether to reduce the number of samples of the target radar signal by several times using the maximum frequency value;
Determining the sampling execution if the sampling reduction ratio is calculated to be equal to or greater than a predetermined ratio;
Determining not to perform sampling for the target radar signal if the sampling reduction rate is less than a predetermined rate
Radar modulation type recognition method.
11. The method of claim 10,
Wherein decreasing the number of samples of the radar signal comprises:
Calculating an average value of a plurality of consecutive samples in the discrete radar signal;
And reducing the number of samples of the discrete radar signal according to the sampling reduction ratio by replacing the plurality of samples with one sample having the calculated average value
Radar modulation type recognition method.
11. The method of claim 10,
The time-
As the time-frequency analysis algorithm, a time-
Radar modulation type recognition method.
11. The method of claim 10,
Wherein performing the image preprocessing comprises:
Reconstructing the image by replacing a brightness value of each pixel on the time-frequency image with a brightness value of each pixel by dividing a brightness value of each pixel by a maximum value of brightness values of the pixels on the time-
Radar modulation type recognition method.
11. The method of claim 10,
Wherein the recognizing comprises:
Inputting a training time-frequency image for a plurality of learning radar signals having different modulation types to a convolutional neural network to learn the convolutional neural network;
And inputting the preprocessed time-frequency image to the learned convolution neural network to recognize the modulated form of the target radar signal
Radar modulation type recognition method.


delete

Images