CN114371474A - Intelligent radar signal sorting method and system based on convolution noise reduction self-encoder - Google Patents

Intelligent radar signal sorting method and system based on convolution noise reduction self-encoder Download PDF

Info

Publication number
CN114371474A
CN114371474A CN202111683638.6A CN202111683638A CN114371474A CN 114371474 A CN114371474 A CN 114371474A CN 202111683638 A CN202111683638 A CN 202111683638A CN 114371474 A CN114371474 A CN 114371474A
Authority
CN
China
Prior art keywords
pulse
radar
encoder
sequence
noise reduction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111683638.6A
Other languages
Chinese (zh)
Inventor
洪淑婕
孙闽红
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Dianzi University
Original Assignee
Hangzhou Dianzi University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Dianzi University filed Critical Hangzhou Dianzi University
Priority to CN202111683638.6A priority Critical patent/CN114371474A/en
Publication of CN114371474A publication Critical patent/CN114371474A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/02—Systems using reflection of radio waves, e.g. primary radar systems; Analogous systems
    • G01S13/06—Systems determining position data of a target
    • G01S13/08—Systems for measuring distance only
    • G01S13/32—Systems for measuring distance only using transmission of continuous waves, whether amplitude-, frequency-, or phase-modulated, or unmodulated
    • G01S13/34—Systems for measuring distance only using transmission of continuous waves, whether amplitude-, frequency-, or phase-modulated, or unmodulated using transmission of continuous, frequency-modulated waves while heterodyning the received signal, or a signal derived therefrom, with a locally-generated signal related to the contemporaneously transmitted signal

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a radar signal intelligent sorting method and system based on a convolution noise reduction self-encoder, wherein the method comprises the following steps: s1, extracting pulse TOA sequences of single radars in the electromagnetic signal database, and sequencing the pulse TOA sequences of each radar sample according to the arrival time to construct aliasing pulse sequences; s2, carrying out pulse coding processing on the pulse TOA sequence and the aliasing pulse sequence of the single radar to obtain a binary coding vector (training data); s3, inputting the obtained binary coding vector into a convolution noise reduction self-encoder for training, wherein the output of the model is the binary coding vector of the target pulse, and storing the trained model; and S4, inputting the received mixed pulse TOA sequence (real data) in the complex electromagnetic environment into a trained convolution noise reduction self-encoder model to complete signal sorting to obtain a target radar pulse sequence. The invention improves the accuracy of radar signal sorting and lays a foundation for further processing radar signal identification, interference and the like.

Description

Intelligent radar signal sorting method and system based on convolution noise reduction self-encoder
Technical Field
The invention belongs to the technical field of radar reconnaissance, and particularly relates to a method and a system for intelligently sorting radar signals based on a convolution noise reduction self-encoder.
Background
The radar signal sorting is a process of classifying radar mixed pulses received by a receiver according to detected radar pulse parameters and certain priori knowledge and according to real radiation sources to which the radar mixed pulses belong. Through radar signal sorting, a plurality of real radar pulse sequences can be obtained, and further processing such as identification, interference and the like is carried out. The radar signal sorting is used as the basis of radar detection signal processing, is an important means for obtaining military information, researches how to eliminate the negative influence of the electromagnetic environment on the radar signal, and has important national defense and military significance.
Currently, radar signal sorting can be roughly divided into two types, the first type is a traditional signal sorting method based on a Pulse Description Word (PDW), and the Pulse Description Word mainly includes Pulse width, arrival direction, carrier frequency, Pulse amplitude and Pulse arrival time. The Pulse Repetition Interval (PRI) is a parameter with a wide variation range and a large number of working patterns in radar signal parameters, and is most widely applied. The method mainly comprises a cumulant difference histogram method, a sequence difference histogram method, a PRI transformation method and a plane transformation method. The second method is a method based on feature extraction and combined machine learning, wherein a transform domain analysis tool is used for firstly carrying out certain form of transformation on signals, more feature parameters are extracted from radar pulse signals for sorting, the reliability of radar signal data is improved, the features mainly comprise stable intra-pulse features such as entropy features, phase-like coefficients and bispectral features, and finally, various classifier models are applied to realize the sorting of the radar signals.
However, the actual battlefield is affected by electromagnetic interference, the signal-to-noise ratio is generally low and there are a lot of TOA estimation errors; on the other hand, the electromagnetic environment is seriously jammed, and the high-density radar pulse is easy to cause the conditions of high pulse spread rate, serious pulse loss and the like of a receiving end, so that the radar rule is damaged to a great extent. Meanwhile, modern radars tend to be diversified in mode and diversified in application, the working state is flexibly switched, pulse groups with different parameter combinations and different time sequence rules can be selected according to actual combat conditions, and the radar has more excellent early warning and detection capabilities than traditional radars. These factors all cause that the traditional radar signal sorting method is difficult to reasonably model and statistically analyze according to the rule of pulse characteristics, and a more serious challenge is brought to signal sorting.
Deep learning, a branch or sub-domain of machine learning, is one of the latest trends of machine learning and artificial intelligence. Deep learning has been widely used in the fields of natural language processing, data mining, computer vision, and autopilot in view of its powerful feature extraction capability. However, the research results of applying the deep learning method to signal sorting are few at present, and if the recognition performance of signal sorting under the complex electromagnetic environment can be effectively improved by using a proper neural network model and a new research idea, the method becomes a new breakthrough in the field of electronic reconnaissance and has important military significance.
Based on the current situation, the invention provides a radar signal intelligent sorting method and system based on a convolution noise reduction self-encoder based on a deep learning theory.
Disclosure of Invention
The invention aims to improve the radar signal sorting performance in a complex electromagnetic environment, and provides a method and a system for intelligently sorting radar signals based on a convolution noise reduction self-encoder.
The advantages of the invention are shown in the following three aspects:
(1) the invention only uses one TOA parameter, thereby effectively shortening the time of feature extraction. Meanwhile, in order to enable the neural network model to process the increasing pulse arrival time as easily as possible, the TOA sequence is converted into binary codes with only 0 and 1, and the linear digitized TOA sequence can accept certain quantization error.
(2) The convolution noise reduction self-encoder used by the invention is greatly superior to the noise reduction self-encoder in space complexity, greatly reduces the number of training parameters, and effectively prevents overfitting of training.
(3) Under the consideration of the parameter changes such as pulse missing rate, staggered pulse rate, TOA estimation error, signal-to-noise ratio and the like and the complex electromagnetic environment with multifunctional radar signals, the sorting accuracy of the invention is obviously superior to that of a PRI conversion method and an SDIF sorting method based on the TOA parameters, and a good foundation is laid for further high-quality processing such as radar signal identification, interference and the like.
In order to achieve the purpose, the invention adopts the following technical scheme:
an intelligent radar signal sorting method based on a convolution noise reduction self-encoder comprises the following steps:
s1, extracting pulse TOA sequences of single radars in the electromagnetic signal database, and sequencing the pulse TOA sequences of each radar sample according to the arrival time to construct aliasing pulse sequences;
s2, respectively carrying out pulse coding processing on the pulse TOA sequence and the aliasing pulse sequence of the single radar in the step S1 to obtain binary coding vectors (training data);
s3, inputting the binary coding vector obtained in the step S2 into a convolution noise reduction self-encoder for training, wherein the output model is the binary coding vector of the target pulse, and the trained model is stored;
and S4, inputting the received mixed pulse TOA sequence (real data) in the complex electromagnetic environment into a trained convolution noise reduction self-encoder model to complete signal sorting to obtain a target radar pulse sequence.
Further, in step S1, a pulse TOA sequence of a single radar in the electromagnetic signal database is extracted, where the TOA sequence of the single radar may be represented as:
T＝{t1,t2,…,ti,…,tN}
wherein, tiThe time of arrival of the ith pulse of a single radar is shown, and N is the number of intercepted pulses.
Further, in step S1, the pulse TOA sequences of each radar sample are sorted according to the arrival time to construct an alias pulse sequence T' ═ { T ═11,t12,t31…,tji,…,tjNThe hybrid TOA sequence of multiple radars can be expressed as:
T'＝{t11,t12,t31…,tji,…,tjN}
wherein, tjiThe time of arrival of the ith pulse of the jth radar is shown, and N shows the number of the jth radar intercepted pulses.
Further, step S2 is specifically: the pulse TOA sequence T of the single radar in step S1 is set to { T }1,t2,…,ti,…,tNAnd aliased pulse sequenceColumn T ═ T11,t12,t31…,tji,…,tjNCarry out pulse coding processing to obtain binary coding vectors x and x
Figure BDA0003438738570000022
(training data); the pulse coding process comprises the following steps:
Figure BDA0003438738570000021
wherein, tunitRepresenting a given unit, the pulse sequence T being based on TunitLinear digitalization is carried out to make it at [0, M.tunit]In the range, M is the total number of units of the pulse sequence, and i is 0,1, … M-1. If the pulse is received in the time window, the vector element corresponding to the time window is represented as 1; if no pulse is received within the time window, its corresponding vector element is represented as 0. M can be represented as:
M＝tNtunit。
further, step S3 is specifically: the resulting binary code vectors x and
Figure BDA0003438738570000035
inputting the convolution noise reduction self-encoder for training, wherein the output of the model is a binary coding vector z of the target pulse, and storing the trained model.
Binary-coded vectors x and
Figure BDA0003438738570000031
inputting the data into a convolutional noise reduction self-encoder for model training, wherein the encoding process and the decoding process of the convolutional noise reduction self-encoder are as follows:
Figure BDA0003438738570000032
z＝σ'(W2y+b2)
where σ and σ' are ReLU activation functions；b1And b2Is a bias vector; w1And W2Is a weight matrix; x is the original data of the image data,
Figure BDA0003438738570000033
data contaminated by noise; y is a hidden layer feature; z is the output layer data.
The minimum loss and high precision of the model are ensured by adjusting network parameters, and because the input x and the output z of the model are binary coding sequences consisting of 0 and 1, the binary cross entropy is selected as an objective function:
Figure BDA0003438738570000034
further, in step S4, the hybrid pulse TOA sequence (real data) received in the complex electromagnetic environment is input into the trained convolutional de-noising self-encoder model to complete signal sorting, so as to obtain the target radar pulse sequence.
The invention also discloses a radar signal intelligent sorting system based on the convolution noise reduction self-encoder, which comprises the following modules:
an aliasing pulse sequence construction module: extracting pulse TOA sequences of single radars in an electromagnetic signal database, and sequencing the pulse TOA sequences of each radar sample according to arrival time to construct aliasing pulse sequences;
a binary coded vector acquisition module: pulse coding processing is carried out on a pulse TOA sequence and an aliasing pulse sequence of a single radar in an aliasing pulse sequence construction module to obtain a binary coding vector;
a model output module: inputting the binary coding vector obtained by the binary coding vector obtaining module into a convolution noise reduction self-encoder for training, wherein the output model is the binary coding vector of the target pulse, and storing the trained model;
a signal sorting module: and inputting the received mixed pulse TOA sequence in a complex electromagnetic environment into a trained convolution noise reduction self-encoder model to complete signal sorting, so as to obtain a target radar pulse sequence.
Preferably, the aliasing pulse sequence construction module extracts the pulse TOA sequence T ═ T of a single radar in the electromagnetic signal database1,t2,…,ti,…,tNThe TOA sequence of a single radar is expressed as:
T＝{t1,t2,…,ti,…,tN}
wherein, tiThe time of arrival of the ith pulse of the radar is shown, and N is the number of intercepted pulses.
Preferably, the aliasing pulse sequence constructing module sequences the pulse TOA sequence of each radar sample according to the arrival time sequence to construct the aliasing pulse sequence T' ═ { T ═ T11,t12,t31…,tji,…,tjNThe mixed TOA sequence of multiple radars is expressed as:
T'＝{t11,t12,t31…,tji,…,tjN}
wherein, tjiThe time of arrival of the ith pulse of the jth radar is shown, and N shows the number of the jth radar intercepted pulses.
Preferably, in the binary coded vector acquisition module, the pulse TOA sequence T of the single radar is T ═ { T ═ T1,t2,…,ti,…,tNT and the alias pulse sequence T ═ T11,t12,t31…,tji,…,tjNCarry out pulse coding processing to obtain binary coding vectors x and x
Figure BDA0003438738570000045
The pulse coding processing process comprises the following steps:
Figure BDA0003438738570000041
wherein, tunitRepresenting a given unit, the pulse sequence T being based on TunitLinear digitalization is carried out to make it at [0, M.tunit]In the range, M is the total number of units of the pulse sequence, i is 0,1, … M-1; if in the time windowIf the pulse is received in the port, the vector element corresponding to the time window is represented as 1; if no pulse is received within the time window, its corresponding vector element is represented as 0; m is expressed as:
M＝tN/tunit。
preferably, the model output module is specifically as follows: inputting binary code vector x sum with size of 1 × 784
Figure BDA0003438738570000046
As the input of the convolution noise reduction self-encoder, the convolution noise reduction self-encoder converts the input size into 28 multiplied by 1 and carries out model training; the encoder part in the convolutional noise reduction self-encoder consists of two convolutional layers and two pooling layers, and the decoder part consists of two anti-convolutional layers and two up-sampling layers; the encoding process and the decoding process of the convolutional noise reduction self-encoder are as follows:
Figure BDA0003438738570000042
z＝σ'(W2y+b2)
wherein σ and σ' are ReLU activation functions; b1And b2Is a bias vector; w1And W2Is a weight matrix; x is the original data of the image data,
Figure BDA0003438738570000043
data contaminated by noise; y is a hidden layer feature; z is output layer data;
the minimum loss and high precision of the model are ensured by adjusting network parameters, and because the input x and the output z of the model are binary coding sequences consisting of 0 and 1, the binary cross entropy is selected as an objective function:
Figure BDA0003438738570000044
compared with the prior art, the method comprehensively considers the influence of a complex electromagnetic environment, and aims at the technical problems that the existing radar signal sorting method considers the parameter changes such as pulse missing rate, staggered pulse rate, TOA estimation error, signal to noise ratio and the like and the sorting performance is reduced in the complex electromagnetic environment with multifunctional radar signals, the arrival time of the pulse sequence is coded and converted into a binary coding vector, the coding vector is input into a convolution noise reduction self-encoder to learn the internal time mode of the target pulse sequence, the mixed pulse sequence is sorted by using a trained network, and the target pulse sequence is extracted.
Drawings
FIG. 1 is a flowchart illustrating a method for sorting radar signals based on a convolutional noise reduction auto-encoder according to an embodiment;
FIG. 2 is a schematic diagram of an exemplary database data processing of electromagnetic signals;
FIG. 3 is a schematic diagram of a convolutional denoising autoencoder model according to an embodiment;
FIG. 4 is a diagram illustrating signal sorting results for different SNR according to an embodiment;
FIG. 5 is a graph illustrating signal sorting results for different pulse loss rates according to one embodiment;
FIG. 6 is a diagram illustrating signal sorting results for different pulse spread rates according to an embodiment;
FIG. 7 is a diagram illustrating signal sorting results for different TOA estimation errors according to an embodiment;
FIG. 8 is a diagram illustrating signal sorting results at different signal-to-noise ratios in an electromagnetic environment with a multifunctional signal according to an embodiment;
FIG. 9 is a diagram illustrating the sorting results of signals under different TOA estimation errors in an electromagnetic environment with a multifunctional signal according to an embodiment;
FIG. 10 is a block diagram of a radar signal sorting system based on a convolution noise reduction self-encoder according to a second embodiment.
Detailed Description
The implementation of the present invention is described in the following preferred embodiments, and other advantages and effects of the present invention can be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention.
Example one
Referring to fig. 1, the embodiment provides an intelligent radar signal sorting method based on a convolution noise reduction self-encoder, which specifically includes the following steps:
s1, extracting pulse TOA sequences of single radars in the electromagnetic signal database, and then sequencing the pulse TOA sequences of each radar sample according to the arrival time to construct aliasing pulse sequences;
s2, respectively carrying out pulse coding processing on the pulse TOA sequence and the aliasing pulse sequence of the single radar in the step S1 to obtain binary coding vectors (training data);
s3, inputting the binary coding vector obtained in the step S2 into a convolution noise reduction self-encoder for training, wherein the output of the model is the binary coding vector of the target pulse, and a trained model is obtained;
and S4, inputting the received mixed pulse TOA sequence (real data) in the complex electromagnetic environment into a trained convolution noise reduction self-encoder model to complete signal sorting to obtain a target radar pulse sequence.
Specifically, in step S1, extracting the pulse TOA sequences of a single radar in the electromagnetic signal database, and then sequencing the pulse TOA sequences of each radar sample according to the sequence of the arrival time to construct an aliasing pulse sequence; in this embodiment, data in the electromagnetic signal database is extracted and pre-processed.
Firstly, extracting a pulse TOA sequence T ═ T of a single radar in an electromagnetic signal database1,t2,…,ti,…,tNThe TOA sequence of a single radar can be expressed as:
T＝{t1,t2,…,ti,…,tN}
wherein, tiThe time of arrival of the ith pulse of the radar is shown, and N is the number of intercepted pulses.
The TOA sequence of pulses for each radar sample is then determined according to its time of arrivalSequentially ordering and constructing an aliasing pulse sequence T ═ T11,t12,t31…,tji,…,tjNReferring to fig. 2, the hybrid TOA sequence of multiple radars can be expressed as:
T'＝{t11,t12,t31…,tji,…,tjN}
wherein, tjiThe time of arrival of the ith pulse of the jth radar is shown, and N shows the number of the jth radar intercepted pulses.
In step S2, the pulse TOA sequence and the alias pulse sequence of the single radar in step S1 are pulse-coded to obtain a binary code vector (training data).
In the present embodiment, data is pulse-coded.
As the pulse arrival time increases, the TOA sequence is finally increased to a larger value, and in order to make the neural network model process the input as easily as possible, the TOA sequence T ═ { T } of the single radar s (T) is determined1,t2,…,ti,…,tNCarrying out pulse coding processing to obtain a binary coding vector x, wherein the pulse coding process comprises the following steps:
Figure BDA0003438738570000061
wherein, tunitRepresenting a given unit, the pulse sequence T being based on TunitLinear digitalization is carried out to make it at [0, M.tunit]In the range, M is the total number of units of the pulse sequence, and i is 0,1, … M-1. If the pulse is received in the time window, the vector element corresponding to the time window is represented as 1; if no pulse is received within the time window, its corresponding vector element is represented as 0. M can be represented as:
M＝tN/tunit。
in step S3, the obtained binary code vector is input to a convolutional noise reduction self-encoder and trained, the output of the model is the binary code vector of the target pulse, and the trained model is stored.
In this embodiment, the model is trained.
Considering that the convolutional noise reduction self-encoder is superior to the noise reduction self-encoder in space complexity, the number of training parameters is greatly reduced, and overfitting of training is effectively prevented, so that the convolutional noise reduction self-encoder is selected as a model to be trained.
Inputting binary code vector x sum with size of 1 × 784
Figure BDA0003438738570000062
As an input to the convolutional noise reduction auto-encoder, the convolutional noise reduction auto-encoder converts the input size to 28 × 28 × 1 and performs model training. The encoder part of the convolutional noise reduction self-encoder consists of two convolutional layers and two pooling layers, and the decoder part consists of two anti-convolutional layers and two up-sampling layers. The specific structure of the network is shown in fig. 3. The encoding process and the decoding process of the convolutional noise reduction self-encoder are as follows:
Figure BDA0003438738570000063
z＝σ'(W2y+b2)
wherein σ and σ' are ReLU activation functions; b1And b2Is a bias vector; w1And W2Is a weight matrix; x is the original data of the image data,
Figure BDA0003438738570000064
data contaminated by noise; y is a hidden layer feature; z is the output layer data.
The purpose of model training is to ensure the minimization of model loss and higher accuracy by adjusting network parameters. Thus, the trained convolutional noise-reducing self-encoder can accurately sort out the target pulse sequence step by step. Due to the input x and output of the model
Figure BDA0003438738570000065
Both are binary coding sequences consisting of 0 and 1, so that binary cross entropy is selected asFor the objective function:
Figure BDA0003438738570000071
in step S4, the hybrid pulse TOA sequence (real data) received in the complex electromagnetic environment is input into the trained convolutional noise reduction self-encoder model to complete signal sorting, so as to obtain a target radar pulse sequence.
In this embodiment, pulse sorting is performed.
And inputting the received mixed pulse TOA sequence (real data) in a complex electromagnetic environment into a trained convolutional noise reduction self-encoder model to complete signal sorting to obtain a target radar pulse sequence. The sorting results of the signals under different electromagnetic environmental conditions are shown in fig. 4 to 9.
The method comprehensively considers the problem of poor radar signal sorting performance in the complex electromagnetic environment, can improve the radar signal sorting accuracy by adding the deep learning algorithm, and lays a foundation for further processing of radar signal identification, interference and the like.
Example two
As shown in fig. 10, the present embodiment discloses an intelligent radar signal sorting system based on a convolution noise reduction self-encoder, which includes the following modules:
an aliasing pulse sequence construction module: extracting pulse TOA sequences of single radars in an electromagnetic signal database, and sequencing the pulse TOA sequences of each radar sample according to arrival time to construct aliasing pulse sequences;
a binary coded vector acquisition module: pulse coding processing is carried out on a pulse TOA sequence and an aliasing pulse sequence of a single radar in an aliasing pulse sequence construction module to obtain a binary coding vector;
a model output module: inputting the binary coding vector obtained by the binary coding vector obtaining module into a convolution noise reduction self-encoder for training, wherein the output model is the binary coding vector of the target pulse, and storing the trained model;
a signal sorting module: and inputting the received mixed pulse TOA sequence in a complex electromagnetic environment into a trained convolution noise reduction self-encoder model to complete signal sorting, so as to obtain a target radar pulse sequence.
In this embodiment, the aliasing pulse sequence constructing module extracts the pulse TOA sequence T ═ T of a single radar in the electromagnetic signal database1,t2,…,ti,…,tNThe TOA sequence of a single radar is expressed as:
T＝{t1,t2,…,ti,…,tN}
wherein, tiThe time of arrival of the ith pulse of the radar is shown, and N is the number of intercepted pulses.
In this embodiment, the aliasing pulse sequence constructing module sequences the pulse TOA sequence of each radar sample according to the arrival time to construct an aliasing pulse sequence T' ═ { T ═ T11,t12,t31…,tji,…,tjNThe mixed TOA sequence of multiple radars is expressed as:
T'＝{t11,t12,t31…,tji,…,tjN}
wherein, tjiThe time of arrival of the ith pulse of the jth radar is shown, and N shows the number of the jth radar intercepted pulses.
In this embodiment, in the binary coded vector obtaining module, the pulse TOA sequence T of the single radar is set to { T ═ T }1,t2,…,ti,…,tNT and the alias pulse sequence T ═ T11,t12,t31…,tji,…,tjNCarry out pulse coding processing to obtain binary coding vectors x and x
Figure BDA0003438738570000081
The pulse coding processing process comprises the following steps:
Figure BDA0003438738570000082
wherein, tunitRepresenting a given unit, a sequence of pulsesT is according to TunitLinear digitalization is carried out to make it at [0, M.tunit]In the range, M is the total number of units of the pulse sequence, i is 0,1, … M-1; if the pulse is received in the time window, the vector element corresponding to the time window is represented as 1; if no pulse is received within the time window, its corresponding vector element is represented as 0; m is expressed as:
M＝tN/tunit。
in this embodiment, the model output module specifically includes: inputting binary code vector x sum with size of 1 × 784
Figure BDA0003438738570000086
As the input of the convolution noise reduction self-encoder, the convolution noise reduction self-encoder converts the input size into 28 multiplied by 1 and carries out model training; the encoder part in the convolutional noise reduction self-encoder consists of two convolutional layers and two pooling layers, and the decoder part consists of two anti-convolutional layers and two up-sampling layers; the encoding process and the decoding process of the convolutional noise reduction self-encoder are as follows:
Figure BDA0003438738570000083
z＝σ'(W2y+b2)
wherein σ and σ' are ReLU activation functions; b1And b2Is a bias vector; w1And W2Is a weight matrix; x is the original data of the image data,
Figure BDA0003438738570000084
data contaminated by noise; y is a hidden layer feature; z is output layer data;
the minimum loss and high precision of the model are ensured by adjusting network parameters, and because the input x and the output z of the model are binary coding sequences consisting of 0 and 1, the binary cross entropy is selected as an objective function:
Figure BDA0003438738570000085
aiming at the problem of poor performance of the traditional radar signal sorting method in a complex electromagnetic environment, the accuracy of radar signal sorting can be improved by adding a deep learning algorithm, and a foundation is laid for further processing radar signal identification, interference and the like.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims (10)

1. The radar signal intelligent sorting method based on the convolution noise reduction self-encoder is characterized by comprising the following steps of:
s1, extracting pulse TOA sequences of single radars in the electromagnetic signal database, and sequencing the pulse TOA sequences of each radar sample according to arrival time to construct aliasing pulse sequences;
s2, carrying out pulse coding processing on the pulse TOA sequence and the aliasing pulse sequence of the single radar in the step S1 to obtain a binary coding vector;
s3, inputting the binary coding vector obtained in the step S2 into a convolution noise reduction self-encoder for training, wherein the output model is the binary coding vector of the target pulse, and the trained model is stored;
and S4, inputting the received mixed pulse TOA sequence in a complex electromagnetic environment into a trained convolutional noise reduction self-encoder model to complete signal sorting, and obtaining a target radar pulse sequence.
2. The intelligent radar signal sorting method based on convolution noise reduction self-encoder according to claim 1,
step S1 extracts the pulse TOA sequence T ═ T of a single radar in the electromagnetic signal database1,t2,…,ti,…,tNThe TOA sequence of a single radar is expressed as:
T＝{t1,t2,…,ti,…,tN}
wherein, tiThe time of arrival of the ith pulse of the radar is shown, and N is the number of intercepted pulses.
3. The intelligent radar signal sorting method based on convolution noise reduction self-encoder according to claim 2,
step S1 sequences the pulse TOA sequence of each radar sample according to the sequence of the arrival time to construct an alias pulse sequence T' ═ { T ═ T11,t12,t31…,tji,…,tjNThe mixed TOA sequence of multiple radars is expressed as:
T'＝{t11,t12,t31…,tji,…,tjN}
wherein, tjiThe time of arrival of the ith pulse of the jth radar is shown, and N shows the number of the jth radar intercepted pulses.
4. The intelligent radar signal sorting method based on convolution noise reduction self-encoder according to claim 3,
in step S2, the pulse TOA sequence T of the single radar is set to { T ═ T }1,t2,…,ti,…,tNT and the alias pulse sequence T ═ T11,t12,t31…,tji,…,tjNCarry out pulse coding processing to obtain binary coding vectors x and x
Figure FDA0003438738560000011
The pulse coding processing process comprises the following steps:
Figure FDA0003438738560000012
wherein, tunitRepresenting a given unit, the pulse sequence T being based on TunitLinear digitalization is carried out to make it at [0, M.tunit]In the range, M is the total number of units of the pulse sequence, i is 0,1, … M-1; if the pulse is received in the time window, the vector element corresponding to the time window is represented as 1; if no pulse is received within the time window, its corresponding vector element is represented as 0; m is expressed as:
M＝tN/tunit。
5. the intelligent radar signal sorting method based on convolution noise reduction self-encoder according to claim 4,
step S3, the obtained binary code vector x and
Figure FDA0003438738560000013
inputting a convolution noise reduction self-encoder for training, wherein the output of the model is a binary coding vector z of the target pulse, and storing the trained model; the method comprises the following specific steps:
inputting binary code vector x sum with size of 1 × 784
Figure FDA0003438738560000021
As the input of the convolution noise reduction self-encoder, the convolution noise reduction self-encoder converts the input size into 28 multiplied by 1 and carries out model training; the encoder part in the convolutional noise reduction self-encoder consists of two convolutional layers and two pooling layers, and the decoder part consists of two anti-convolutional layers and two up-sampling layers; the encoding process and the decoding process of the convolutional noise reduction self-encoder are as follows:
Figure FDA0003438738560000022
z＝σ'(W2y+b2)
wherein σ and σ' are ReLU activation functions; b1And b2Is a bias vector; w1And W2Is a weight matrix; x is the original data of the image data,
Figure FDA0003438738560000023
data contaminated by noise; y is a hidden layer feature; z is output layer data;
the minimum loss and high precision of the model are ensured by adjusting network parameters, and because the input x and the output z of the model are binary coding sequences consisting of 0 and 1, the binary cross entropy is selected as an objective function:
Figure FDA0003438738560000024
6. radar signal intelligence system of sorting based on convolutional self-encoder of making an uproar that falls, its characterized in that includes the following module:
an aliasing pulse sequence construction module: extracting pulse TOA sequences of single radars in an electromagnetic signal database, and sequencing the pulse TOA sequences of each radar sample according to arrival time to construct aliasing pulse sequences;
a binary coded vector acquisition module: pulse coding processing is carried out on a pulse TOA sequence and an aliasing pulse sequence of a single radar in an aliasing pulse sequence construction module to obtain a binary coding vector;
a model output module: inputting the binary coding vector obtained by the binary coding vector obtaining module into a convolution noise reduction self-encoder for training, wherein the output model is the binary coding vector of the target pulse, and storing the trained model;
a signal sorting module: and inputting the received mixed pulse TOA sequence in a complex electromagnetic environment into a trained convolution noise reduction self-encoder model to complete signal sorting, so as to obtain a target radar pulse sequence.
7. The convolution noise reduction self-encoder based radar signal intelligent sorting system of claim 6,
an aliasing pulse sequence construction module extracts a pulse TOA sequence T ═ T of a single radar in an electromagnetic signal database1,t2,…,ti,…,tNThe TOA sequence of a single radar is expressed as:
T＝{t1,t2,…,ti,…,tN}
wherein, tiThe time of arrival of the ith pulse of the radar is shown, and N is the number of intercepted pulses.
8. The convolution noise reduction self-encoder based radar signal intelligent sorting system of claim 7,
the aliasing pulse sequence construction module sequences the pulse TOA sequence of each radar sample according to the arrival time sequence thereof to construct an aliasing pulse sequence T' ═ { T }11,t12,t31…,tji,…,tjNThe mixed TOA sequence of multiple radars is expressed as:
T'＝{t11,t12,t31…,tji,…,tjN}
wherein, tjiThe time of arrival of the ith pulse of the jth radar is shown, and N shows the number of the jth radar intercepted pulses.
9. The convolution noise reduction self-encoder based radar signal intelligent sorting system of claim 8,
in the binary code vector acquisition module, the pulse TOA sequence T ═ T of a single radar1,t2,…,ti,…,tNT and the alias pulse sequence T ═ T11,t12,t31…,tji,…,tjNCarry out pulse coding processing to obtain binary coding vectors x and x
Figure FDA0003438738560000031
The pulse coding processing process comprises the following steps:
Figure FDA0003438738560000032
wherein, tunitRepresenting a given unit, the pulse sequence T being based on TunitLinear digitalization is carried out to make it at [0, M.tunit]In the range, M is the total number of units of the pulse sequence, i is 0,1, … M-1; if the pulse is received in the time window, the vector element corresponding to the time window is represented as 1; if no pulse is received within the time window, its corresponding vector element is represented as 0; m is expressed as:
M＝tN/tunit。
10. the convolution noise reduction self-encoder based radar signal intelligent sorting system of claim 9,
the model output module is specifically as follows: inputting binary code vector x sum with size of 1 × 784
Figure FDA0003438738560000033
As the input of the convolution noise reduction self-encoder, the convolution noise reduction self-encoder converts the input size into 28 multiplied by 1 and carries out model training; the encoder part in the convolutional noise reduction self-encoder consists of two convolutional layers and two pooling layers, and the decoder part consists of two anti-convolutional layers and two up-sampling layers; the encoding process and the decoding process of the convolutional noise reduction self-encoder are as follows:
Figure FDA0003438738560000034
z＝σ'(W2y+b2)
wherein σ and σ' are ReLU activation functions; b1And b2Is a bias vector; w1And W2Is a weight matrix; x is the original data of the image data,
Figure FDA0003438738560000035
data contaminated by noise; y is a hidden layer feature; z is output layer data;
the minimum loss and high precision of the model are ensured by adjusting network parameters, and because the input x and the output z of the model are binary coding sequences consisting of 0 and 1, the binary cross entropy is selected as an objective function:
Figure FDA0003438738560000036
CN202111683638.6A 2021-12-28 2021-12-28 Intelligent radar signal sorting method and system based on convolution noise reduction self-encoder Pending CN114371474A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111683638.6A CN114371474A (en) 2021-12-28 2021-12-28 Intelligent radar signal sorting method and system based on convolution noise reduction self-encoder

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111683638.6A CN114371474A (en) 2021-12-28 2021-12-28 Intelligent radar signal sorting method and system based on convolution noise reduction self-encoder

Publications (1)

Publication Number Publication Date
CN114371474A true CN114371474A (en) 2022-04-19

Family

ID=81141546

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111683638.6A Pending CN114371474A (en) 2021-12-28 2021-12-28 Intelligent radar signal sorting method and system based on convolution noise reduction self-encoder

Country Status (1)

Country Link
CN (1) CN114371474A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115598600A (en) * 2022-11-28 2023-01-13 四川九洲电器集团有限责任公司（Cn） Secondary radar signal dynamic coding system, method, electronic equipment and medium
CN116774154A (en) * 2023-08-23 2023-09-19 吉林大学 Radar signal sorting method

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115598600A (en) * 2022-11-28 2023-01-13 四川九洲电器集团有限责任公司（Cn） Secondary radar signal dynamic coding system, method, electronic equipment and medium
CN115598600B (en) * 2022-11-28 2023-03-28 四川九洲电器集团有限责任公司 Secondary radar signal dynamic coding system, method, electronic equipment and medium
CN116774154A (en) * 2023-08-23 2023-09-19 吉林大学 Radar signal sorting method
CN116774154B (en) * 2023-08-23 2023-10-31 吉林大学 Radar signal sorting method

Similar Documents

Publication Publication Date Title
CN113156391B (en) Radar signal multi-dimensional feature intelligent sorting method
CN107220606B (en) Radar radiation source signal identification method based on one-dimensional convolutional neural network
Li et al. Toward convolutional neural networks on pulse repetition interval modulation recognition
CN108416290B (en) Radar signal feature extraction method based on residual deep learning
CN110109060B (en) Radar radiation source signal sorting method and system based on deep learning network
CN109683161B (en) Inverse synthetic aperture radar imaging method based on depth ADMM network
CN114371474A (en) Intelligent radar signal sorting method and system based on convolution noise reduction self-encoder
CN112098957A (en) Complex radar radiation source identification method based on one-dimensional self-walking convolution neural network
CN110133599B (en) Intelligent radar radiation source signal classification method based on long-time and short-time memory model
Pardo et al. Seismic phase picking using convolutional networks
CN106483514B (en) Airplane motion mode identification method based on EEMD and support vector machine
CN112684427B (en) Radar target recognition method based on serial secondary reinforcement training
CN113657491A (en) Neural network design method for signal modulation type recognition
CN113759323B (en) Signal sorting method and device based on improved K-Means joint convolution self-encoder
CN112597820A (en) Target clustering method based on radar signal sorting
Liu Pulse deinterleaving for multifunction radars with hierarchical deep neural networks
Guo et al. Radar target recognition based on feature Pyramid fusion lightweight CNN
Zhang et al. VGM-RNN: HRRP sequence extrapolation and recognition based on a novel optimized RNN
CN116797796A (en) Signal identification method based on time-frequency analysis and deep learning under DRFM intermittent sampling
Huang et al. Radar waveform recognition based on multiple autocorrelation images
Zhao et al. Research on electronic jamming identification based on CNN
Kong et al. A transformer-based CTDNN structure for automatic modulation recognition
CN111488552A (en) Close-proximity multi-target tracking method based on Gaussian mixture probability hypothesis density
Ruan et al. Automatic recognition of radar signal types based on CNN-LSTM
Guven et al. Classifying LPI radar waveforms with time-frequency transformations using multi-stage CNN system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination