CN117289218B - Active interference identification method based on attention cascade network - Google Patents
Active interference identification method based on attention cascade network Download PDFInfo
- Publication number
- CN117289218B CN117289218B CN202311581938.2A CN202311581938A CN117289218B CN 117289218 B CN117289218 B CN 117289218B CN 202311581938 A CN202311581938 A CN 202311581938A CN 117289218 B CN117289218 B CN 117289218B
- Authority
- CN
- China
- Prior art keywords
- interference
- network
- identification
- layer
- attention
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000010586 diagram Methods 0.000 claims abstract description 26
- 230000002159 abnormal effect Effects 0.000 claims abstract description 6
- 238000002592 echocardiography Methods 0.000 claims abstract description 6
- 238000000605 extraction Methods 0.000 claims description 20
- 238000011176 pooling Methods 0.000 claims description 10
- 230000007246 mechanism Effects 0.000 claims description 9
- 238000001228 spectrum Methods 0.000 claims description 7
- 239000006185 dispersion Substances 0.000 claims description 5
- 238000003384 imaging method Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 230000002787 reinforcement Effects 0.000 claims 2
- 230000000694 effects Effects 0.000 abstract description 5
- 239000010410 layer Substances 0.000 description 45
- 230000001629 suppression Effects 0.000 description 6
- 230000006870 function Effects 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 239000002355 dual-layer Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
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
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/36—Means for anti-jamming, e.g. ECCM, i.e. electronic counter-counter measures
-
- 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
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/021—Auxiliary means for detecting or identifying radar signals or the like, e.g. radar jamming signals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to an active interference identification method based on an attention cascade network, which comprises the following steps: acquiring a one-dimensional time domain interference signal of an interference signal to be detected; obtaining a time-frequency characteristic diagram according to the one-dimensional time-domain interference signal; inputting the time-frequency characteristic diagram into a double-layer cascade identification network after training, wherein the double-layer cascade identification network comprises a first identification network and a second identification network which are cascaded; obtaining a first identification result by using a first identification network, and outputting the first identification result if the first identification result belongs to a first interference type; if the first identification result is an abnormal result, executing the next step; accumulating echoes of the interference signals to be detected within the whole synthetic aperture time to obtain corresponding two-dimensional interference echo signals; obtaining two-dimensional image information according to the two-dimensional interference echo signals; and inputting the two-dimensional image information into a second recognition network to obtain a second recognition result, wherein the second recognition result belongs to a second interference type. The method provided by the invention has the advantages of more interference types and better effect.
Description
Technical Field
The invention belongs to the technical field of interference identification, and particularly relates to an active interference identification method based on an attention cascade network.
Background
Radar is an important tool for electronic countermeasure, and as the level of electronic countermeasure is continuously improved, interference technology to radar is also increasing. The accurate and stable work of the radar is hindered by various interference signals, and the improvement of the anti-interference technology of the radar is the first problem to be solved by electronic countermeasure. The sensing of the interference signal is the first step of the anti-interference technology, the kind of the interference signal can be sensed and identified, and the interference can be effectively resisted.
The existing active interference recognition technology mainly utilizes selected characteristics to extract characteristics of interference signals, and then utilizes different classifiers to classify the interference signals. The time domain feature extraction is to extract features by using one-dimensional echo time domain signals, and common time domain features include time domain skewness coefficients, time domain kurtosis coefficients, time domain margin, zero center normalized amplitude standard deviation and box dimension. The time domain signal can be converted into a frequency domain for feature extraction, and common frequency domain features include frequency domain skewness coefficient, frequency domain kurtosis coefficient and frequency spectrum envelope waviness. The time-frequency transformation tool can be used for transforming the interfered echo signals into a time-frequency domain, the time-frequency tool is used for wavelet transformation, short-time Fourier transformation, smooth pseudo Wigner-Willi distribution and the like, and the characteristics selectable in the time-frequency domain are singular value entropy, standard deviation of a time-frequency image and characteristic extraction of a gray level co-occurrence matrix of the time-frequency image. After the feature extraction results of the interference signals in different domains are obtained, the interference signals are classified by using a classifier, and the most common classifier is an SVM (Support Vector Machine ) classifier.
However, in the prior art, the extraction of the interference features mainly depends on the specified feature values, and the selection of the feature values is manually selected, so that the accuracy and stability of identification are difficult to ensure. The recognition effect is difficult to predict for a change in the form of the disturbance or a change in the disturbance parameter.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides an active interference identification method based on an attention cascade network. The technical problems to be solved by the invention are realized by the following technical scheme:
the invention provides an active interference identification method based on an attention cascade network, which comprises the following steps:
step 1: acquiring a one-dimensional time domain interference signal of an interference signal to be detected;
step 2: obtaining a time-frequency characteristic diagram according to the one-dimensional time-domain interference signal;
step 3: inputting the time-frequency characteristic diagram into a trained double-layer cascade identification network based on an attention mechanism, wherein the double-layer cascade identification network comprises a first identification network and a second identification network which are cascaded;
step 4: obtaining a first identification result by using the first identification network, and outputting the first identification result if the first identification result belongs to a first interference type; if the first identification result is an abnormal result, executing the step 5;
step 5: accumulating echoes of the interference signals to be detected in the whole synthetic aperture time to obtain corresponding two-dimensional interference echo signals;
step 6: obtaining two-dimensional image information according to the two-dimensional interference echo signals;
step 7: and inputting the two-dimensional image information into the second recognition network to obtain a second recognition result, wherein the second recognition result belongs to a second interference type.
Compared with the prior art, the invention has the beneficial effects that:
the active interference identification method based on the attention cascade network utilizes the double-layer cascade identification network based on the attention mechanism to realize the identification of active interference, utilizes the first identification network to identify the types of suppression type interference, utilizes the image domain information of interference signals in the second identification network, uses the residual network based on the attention mechanism, and more efficiently and accurately identifies the types of dense false target type interference on the premise of effectively preventing the network from being over-fitted.
The foregoing description is only an overview of the present invention, and is intended to be implemented in accordance with the teachings of the present invention, as well as the preferred embodiments thereof, together with the following detailed description of the invention, given by way of illustration only, together with the accompanying drawings.
Drawings
Fig. 1 is a schematic diagram of an active interference identification method based on an attention cascade network according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a first identification network according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a second identification network according to an embodiment of the present invention;
fig. 4 is a schematic diagram of an SE attention module according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following describes an active interference identification method based on the attention cascade network according to the invention in detail with reference to the attached drawings and the detailed description.
The foregoing and other features, aspects, and advantages of the present invention will become more apparent from the following detailed description of the preferred embodiments when taken in conjunction with the accompanying drawings. The technical means and effects adopted by the present invention to achieve the intended purpose can be more deeply and specifically understood through the description of the specific embodiments, however, the attached drawings are provided for reference and description only, and are not intended to limit the technical scheme of the present invention.
Referring to fig. 1, fig. 1 is a schematic diagram of an active interference identification method based on an attention cascade network according to an embodiment of the present invention, as shown in fig. 1, the active interference identification method based on an attention cascade network according to an embodiment of the present invention includes:
step 1: acquiring a one-dimensional time domain interference signal of an interference signal to be detected;
in an alternative embodiment, the interference signal to be detected is obtained, and the interference signal to be detected is extracted along the azimuth direction to obtain the one-dimensional time domain interference signal.
Step 2: obtaining a time-frequency characteristic diagram according to the one-dimensional time-domain interference signal;
in an alternative embodiment, step 2 includes:
step 2.1: performing STFT (short-time Fourier transform short-time Fourier transform) on the one-dimensional time domain interference signal to obtain an original time-frequency characteristic diagram;
step 2.2: and gray scale adjustment is carried out on the original time-frequency characteristic diagram, so that the time-frequency characteristic diagram is obtained.
Optionally, gray scale adjustment is performed on the original time-frequency feature map by using an exponential tool, specifically, an amplitude spectrum of the time-frequency feature map is taken, then the amplitude spectrum is converted into a gray scale image, finally, gray scale adjustment is performed on the gray scale image by using an exponential function, and a gray scale adjusted image, namely, the time-frequency feature map is expressed as follows:
；
wherein,for grey scale image +.>Is a gray scale adjustment coefficient.
Step 3: inputting the time-frequency characteristic diagram into a trained double-layer cascade identification network based on an attention mechanism, wherein the double-layer cascade identification network comprises a first identification network and a second identification network which are cascaded;
step 4: obtaining a first identification result by using a first identification network, and outputting the first identification result if the first identification result belongs to a first interference type; if the first identification result is an abnormal result, executing the step 5;
in an alternative embodiment, the first interference type is a hold-down type interference comprising: radio frequency interference, amplitude modulation interference, phase modulation interference, and frequency modulation interference.
In an alternative embodiment, the anomaly results include a recognized result and an unrecognized result that do not belong to the first interference type.
In an exemplary embodiment, when the classification tag of the first recognition result does not belong to any of the interference classification tags of radio frequency interference, amplitude modulation interference, phase modulation interference and frequency modulation interference, or the classification tag of the first recognition result is error, or the first recognition result does not have any classification tag, the first recognition result is considered to be an abnormal result.
In this embodiment, the first recognition network is configured to extract an image edge distribution feature and an energy distribution feature of the time-frequency feature map, recognize an interference type based on the image edge distribution feature and the energy distribution feature, and output a first recognition result.
Please refer to fig. 2 in combination to a schematic structural diagram of a first identification network provided by an embodiment of the present invention, in which Input is a time-frequency feature diagram, conv is a convolution layer, stride is a step length, and optionally, the first identification network includes a first convolution layer, a first maximum pooling layer Maxpool, a plurality of first feature extraction units, a first average pooling layer avgpool2d, and a first full-connection layer unit that are sequentially cascaded; the first feature extraction unit comprises 2 cascaded first residual modules, and the first residual modules are a 2-layer convolution layer network comprising one shortcut connection; the first full connection layer unit includes 2 full connection layers Fc in cascade.
For example, 4 first feature extraction units may be provided, where the number of output channels of the convolution layers in the 4 first feature extraction units is different, and is 16, 32, 64, and 128, respectively. Wherein, a convolution layer of 1×1 is arranged in the residual module connected with the previous first feature extraction unit to realize channel dimension increase.
Since the statistical characteristics of the suppression type interference in the time-frequency domain can be directly identified by deep learning, the specific types of the suppression type interference, namely, radio frequency interference, amplitude modulation interference, phase modulation interference or frequency modulation interference, can be directly identified after the identification of the first layer identification network of the double-layer cascade identification network. For the interference signals to be detected which do not belong to the suppression type of interference, the second identification network can continue to process to determine the interference type.
Step 5: accumulating echoes of the interference signals to be detected within the whole synthetic aperture time to obtain corresponding two-dimensional interference echo signals;
the second recognition network of the double-layer cascade recognition network is different from the first recognition network, the input of the first recognition network is the time-frequency domain characteristic of a one-dimensional signal, only the time-frequency domain characteristic of a single pulse is needed to be used as the input of the network, the second recognition network utilizes the image domain information of the signal, two-dimensional echoes of the whole imaging scene are needed to be obtained, and the echoes of the interference signal to be detected in the whole synthetic aperture time are needed to be accumulated, so that two-dimensional interference echo signals are obtained.
Step 6: obtaining two-dimensional image information according to the two-dimensional interference echo signals;
in an alternative embodiment, step 6 includes:
step 6.1: imaging the two-dimensional interference echo signals to obtain original two-dimensional image information;
optionally, an RD (Range-Doppler) imaging algorithm is used to image the two-dimensional interference echo signals to obtain original two-dimensional image information.
Step 6.2: and carrying out gray scale adjustment on the original two-dimensional image to obtain two-dimensional image information.
Optionally, the original two-dimensional image is gray-scale adjusted using a logarithmic tool. Two-dimensional image information obtained by imaging algorithmThe gray value at the point is +.>In this embodiment, the gray level image is normalized, and then the gray level value of the gray level image is added integrally to prevent the result of logarithmic operation from being negative, and then the gray level image is subjected to logarithmic operation to obtain the image after gray level adjustment, that is, the two-dimensional image information, after the gray level adjustment is performed on the original two-dimensional image, the intensity of interference is obviously increased, and details are more prominent, so that details of the interference effect are enhanced. The two-dimensional image information is expressed as:
。
step 7: and inputting the two-dimensional image information into a second recognition network to obtain a second recognition result, wherein the second recognition result belongs to a second interference type.
In an alternative embodiment, the second interference type is dense decoy type interference comprising: random frequency shifting interference, fixed frequency shifting interference, spectrum dispersion interference, intermittent sampling interference and micro-motion modulation interference.
It is noted that for dense decoy type interference, the time-frequency domain cannot exhibit statistical characteristics that match the neural network, and therefore further processing using the second identification network is required.
Please refer to fig. 3 in combination to a schematic structural diagram of a second identification network provided by the embodiment of the present invention, in which Input is two-dimensional image information, conv is a convolution layer, stride is a step length, and optionally, the second identification network includes a second convolution layer, a second maximum pooling layer Maxpool, a plurality of second feature extraction units, a second average pooling layer avgpool2d, and a second full connection layer unit that are sequentially cascaded; the second feature extraction unit comprises a second residual error module and a residual error attention module which are connected in cascade, wherein the second residual error module is a 2-layer convolution layer network comprising a shortcut connection (shortcut), and the residual error attention module is a 2-layer convolution layer network comprising an SE (sequential-and-specification) attention module and a shortcut connection; the second full connection layer unit includes 2 full connection layers Fc in cascade.
For example, 3 second feature extraction units may be provided, where the number of output channels of the convolution layers in the 3 second feature extraction units is different, 16, 32, and 64, respectively. Wherein, a convolution layer of 1×1 is arranged in the residual module connected with the previous second feature extraction unit to realize channel dimension increase.
In this embodiment, the residual attention module is formed by adding an SE attention module after the second layer convolutional layer of the second residual module. As shown in the schematic diagram of the SE attention module in fig. 4, H, W, C in the figure represents three dimensions of rows, columns and channel numbers, respectively, the SE attention module in this embodiment includes a global pooling layer, a first fully connected layer, a ReLu function layer, a second fully connected layer and a Sigmoid function layer that are cascaded in sequence.
In this embodiment, the second recognition network is configured to extract an interference energy distribution feature of the two-dimensional image information, perform channel feature enhancement on the interference energy distribution feature by using the SE attention module, recognize an interference type based on the interference energy distribution feature after the channel feature enhancement, and output a second recognition result.
In this embodiment, the SE attention module provides an attention mechanism to make the feature extraction of the region of interest by the second recognition network more concentrated, so that the network assigns a higher weight to the feature of interest. In addition, the residual error module can also enhance the robustness of the network, and the dense false target type interference is identified through the residual error module containing the attention mechanism, so that the specific type of the dense false target type interference is finally identified, namely: random frequency shifting interference, fixed frequency shifting interference, spectrum dispersion interference, intermittent sampling interference or micro-motion modulation interference.
Further, training is required before the interference recognition of the interference signal to be detected by using the dual-layer cascade recognition network. Firstly, a double-layer cascade identification network is established, and an interference signal data set is obtained, wherein the interference signal data set comprises a plurality of interference signals corresponding to a plurality of interference types such as radio frequency interference, amplitude modulation interference, phase modulation interference, frequency modulation interference, random frequency shift interference, fixed frequency shift interference, frequency spectrum dispersion interference, intermittent sampling interference, micro-motion modulation interference and the like. Then, processing the interference signals belonging to the suppression type interference to obtain a corresponding time-frequency characteristic diagram, endowing the time-frequency characteristic diagram with a corresponding interference type label, and forming a first training sample set by the time-frequency characteristic diagram endowed with the interference type label; and processing the interference signals belonging to the dense false target type interference to obtain corresponding two-dimensional image information, endowing the two-dimensional image information with corresponding interference type labels, and endowing the two-dimensional image information with the interference type labels as a second training sample set. The first training sample set and the second training sample set are respectively input into a first recognition network and a second recognition network of the double-layer cascade recognition network for training, so that the double-layer cascade recognition network after training is obtained, and the specific training method is an existing network training method and is not described in detail herein.
In this embodiment, the interference signal in the interference signal data set is obtained by performing simulation interference based on a chirp (linear frequency modulation, LFM) signal. Set the duration of an ideal chirp signal asSecond, amplitude is constant, center frequency is +.>Phase->The LFM signal is expressed as: />Wherein->Is a time variable +.>For the duration of the transmitted signal pulse, +.>For signalling frequency, < >>Is imaginary unit, ++>As a rectangular function.
Radio-frequency noise interference (RF) is usually a narrowband gaussian process, amplitude Modulation noise interference (Amplitude Modulation, AM) is interference that modulates an interference signal in amplitude, phase Modulation noise interference (PM) generates an interference signal by modulating a signal Phase, frequency Modulation noise interference (Frequency Modulation, FM) is noise that modulates a signal in frequency, and the corresponding interference signal is obtained after performing the above-described simulation interference processing on an LFM signal.
For dense false target type interference, firstly, an LFM signal is changed into an intermediate frequency signal through down conversion, then a digital signal is obtained through analog-to-digital conversion, and then the digital signal is subjected to analog signal conversion output under the clock control rate of m times, so that an SMSP (spectral dispersion) signal is obtained. And (3) repeating intermittent sampling forwarding, random frequency shifting, fixed frequency shifting, micro-motion modulation and short-time Fourier transformation processing are carried out on the SMSP signals, and then corresponding interference signals are obtained.
According to the active interference identification method based on the attention cascade network, the identification of active interference is realized by using the double-layer cascade identification network based on the attention mechanism, the types of suppression type interference are identified by using the first-layer identification network, the identification of the types of interference is performed by using the image domain information of interference signals in the second-layer identification network, the residual network based on the attention mechanism is used, and the types of dense false target type interference are identified more efficiently and accurately on the premise of effectively preventing the network from being over-fitted. In addition, the neural network can be used for extracting high-level semantic information of the input signal, so that the application range of the input signal is wider than that of manually selected features, and the stability is higher.
According to the active interference identification method based on the attention cascade network, the overall accuracy of identification is more than 95%, and on the premise that the identification of the interference types is more, the accuracy is improved by 9.98% compared with that of the traditional SVM identification method.
It should be noted that in this document relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that an article or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in an article or apparatus that comprises the element. The terms "connected" or "connected," and the like, are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect.
The foregoing is a further detailed description of the invention in connection with the preferred embodiments, and it is not intended that the invention be limited to the specific embodiments described. It will be apparent to those skilled in the art that several simple deductions or substitutions may be made without departing from the spirit of the invention, and these should be considered to be within the scope of the invention.
Claims (8)
1. An active interference identification method based on an attention cascade network, comprising:
step 1: acquiring a one-dimensional time domain interference signal of an interference signal to be detected;
step 2: obtaining a time-frequency characteristic diagram according to the one-dimensional time-domain interference signal; comprising the following steps:
step 2.1: performing STFT (short-time Fourier transform) on the one-dimensional time domain interference signal to obtain an original time-frequency characteristic diagram;
step 2.2: gray scale adjustment is carried out on the original time-frequency characteristic diagram, and the time-frequency characteristic diagram is obtained;
step 3: inputting the time-frequency characteristic diagram into a trained double-layer cascade identification network based on an attention mechanism, wherein the double-layer cascade identification network comprises a first identification network and a second identification network which are cascaded;
step 4: obtaining a first identification result by using the first identification network, and outputting the first identification result if the first identification result belongs to a first interference type; if the first identification result is an abnormal result, executing the step 5;
step 5: accumulating echoes of the interference signals to be detected in the whole synthetic aperture time to obtain corresponding two-dimensional interference echo signals;
step 6: obtaining two-dimensional image information according to the two-dimensional interference echo signals; comprising the following steps:
step 6.1: imaging the two-dimensional interference echo signals to obtain original two-dimensional image information;
step 6.2: gray scale adjustment is carried out on the original two-dimensional image, so that the two-dimensional image information is obtained;
step 7: and inputting the two-dimensional image information into the second recognition network to obtain a second recognition result, wherein the second recognition result belongs to a second interference type.
2. The method for identifying active interference based on an attention cascade network according to claim 1, wherein the step 1 comprises:
and acquiring the interference signal to be detected, and extracting the interference signal to be detected along the azimuth direction to obtain the one-dimensional time domain interference signal.
3. The method of attention cascade network based active disturbance identification of claim 1, wherein the first disturbance type is a throttle type disturbance, the throttle type disturbance comprising: radio frequency interference, amplitude modulation interference, phase modulation interference, and frequency modulation interference.
4. The method of attention cascade network based active disturbance identification according to claim 1, wherein the abnormal result includes a recognized result and an unrecognized result which do not belong to the first disturbance type.
5. The method for identifying active disturbance based on an attention cascade network according to claim 3,
the first identification network comprises a first convolution layer, a first maximum pooling layer, a plurality of first feature extraction units, a first average pooling layer and a first full-connection layer unit which are sequentially cascaded; the first feature extraction unit comprises 2 cascaded first residual modules, and the first residual modules are a 2-layer convolution layer network comprising one shortcut connection; the first full connection layer unit comprises 2 cascaded full connection layers;
the first recognition network is used for extracting image edge distribution characteristics and energy distribution characteristics of the time-frequency characteristic map, recognizing the interference type based on the image edge distribution characteristics and the energy distribution characteristics, and outputting the first recognition result.
6. The method of attention cascade network based active interference identification of claim 1, wherein the second interference type is dense decoy type interference comprising: random frequency shifting interference, fixed frequency shifting interference, spectrum dispersion interference, intermittent sampling interference and micro-motion modulation interference.
7. The method for active disturbance recognition based on an attention cascade network according to claim 6,
the second identification network comprises a second convolution layer, a second maximum pooling layer, a plurality of second feature extraction units, a second average pooling layer and a second full-connection layer unit which are sequentially cascaded; the second feature extraction unit comprises a second residual error module and a residual error attention module which are connected in cascade, wherein the second residual error module is a 2-layer convolution layer network comprising one shortcut connection, and the residual error attention module is a 2-layer convolution layer network comprising an SE attention module and one shortcut connection; the second full connection layer unit comprises 2 cascaded full connection layers;
the second recognition network is used for extracting interference energy distribution characteristics of the two-dimensional image information, the SE attention module is used for carrying out channel characteristic reinforcement on the interference energy distribution characteristics, the interference type is recognized based on the interference energy distribution characteristics after the channel characteristic reinforcement, and the second recognition result is output.
8. The attention cascade network-based active disturbance identification method of claim 7, wherein the SE attention module comprises a global pooling layer, a first fully-connected layer, a ReLu function layer, a second fully-connected layer, and a Sigmoid function layer, which are cascaded in sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311581938.2A CN117289218B (en) | 2023-11-24 | 2023-11-24 | Active interference identification method based on attention cascade network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311581938.2A CN117289218B (en) | 2023-11-24 | 2023-11-24 | Active interference identification method based on attention cascade network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117289218A CN117289218A (en) | 2023-12-26 |
| CN117289218B true CN117289218B (en) | 2024-02-06 |
Family
ID=89239410
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311581938.2A Active CN117289218B (en) | 2023-11-24 | 2023-11-24 | Active interference identification method based on attention cascade network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117289218B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110927706A (en) * | 2019-12-10 | 2020-03-27 | 电子科技大学 | Convolutional neural network-based radar interference detection and identification method |
| EP3674741A1 (en) * | 2018-12-27 | 2020-07-01 | Avantix | System and method for identifying a radar source |
| CN112859012A (en) * | 2021-01-20 | 2021-05-28 | 北京理工大学 | Radar deception jamming identification method based on cascade convolution neural network |
| CN113920320A (en) * | 2021-08-27 | 2022-01-11 | 北京临近空间飞行器系统工程研究所 | Radar image target detection system for typical active interference |
| CN114580461A (en) * | 2022-01-20 | 2022-06-03 | 西安电子科技大学 | Training method of radar interference recognition network and radar interference recognition method |
| CN115097396A (en) * | 2022-06-21 | 2022-09-23 | 西安电子科技大学 | Radar active interference identification method based on CNN and LSTM series model |
| CN115616503A (en) * | 2022-11-10 | 2023-01-17 | 西北工业大学 | Radar interference signal type identification method based on convolutional neural network model |
| CN116047427A (en) * | 2023-03-29 | 2023-05-02 | 西安电子科技大学 | Small sample radar active interference identification method |
| CN116797796A (en) * | 2023-06-28 | 2023-09-22 | 北京理工大学 | Signal identification method based on time-frequency analysis and deep learning under DRFM intermittent sampling |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11269058B2 (en) * | 2018-06-13 | 2022-03-08 | Metawave Corporation | Autoencoder assisted radar for target identification |
| CN115049952B (en) * | 2022-04-24 | 2023-04-07 | 南京农业大学 | Juvenile fish limb identification method based on multi-scale cascade perception deep learning network |
-
2023
- 2023-11-24 CN CN202311581938.2A patent/CN117289218B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3674741A1 (en) * | 2018-12-27 | 2020-07-01 | Avantix | System and method for identifying a radar source |
| CN110927706A (en) * | 2019-12-10 | 2020-03-27 | 电子科技大学 | Convolutional neural network-based radar interference detection and identification method |
| CN112859012A (en) * | 2021-01-20 | 2021-05-28 | 北京理工大学 | Radar deception jamming identification method based on cascade convolution neural network |
| CN113920320A (en) * | 2021-08-27 | 2022-01-11 | 北京临近空间飞行器系统工程研究所 | Radar image target detection system for typical active interference |
| CN114580461A (en) * | 2022-01-20 | 2022-06-03 | 西安电子科技大学 | Training method of radar interference recognition network and radar interference recognition method |
| CN115097396A (en) * | 2022-06-21 | 2022-09-23 | 西安电子科技大学 | Radar active interference identification method based on CNN and LSTM series model |
| CN115616503A (en) * | 2022-11-10 | 2023-01-17 | 西北工业大学 | Radar interference signal type identification method based on convolutional neural network model |
| CN116047427A (en) * | 2023-03-29 | 2023-05-02 | 西安电子科技大学 | Small sample radar active interference identification method |
| CN116797796A (en) * | 2023-06-28 | 2023-09-22 | 北京理工大学 | Signal identification method based on time-frequency analysis and deep learning under DRFM intermittent sampling |
Non-Patent Citations (2)
| Title |
|---|
| 孙忠镇等.基于级联网络的复杂大场景SAR图像舰船目标快速检测.信号处理.2021,第37卷(第6期),第941-951页. * |
| 熊伟 ; 曹兰英 ; 郝志梅 ; .基于多尺度相像系数的雷达干扰类型频域识别.计算机仿真.2010,(第03期),第19-22页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117289218A (en) | 2023-12-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Work modes recognition and boundary identification of MFR pulse sequences with a hierarchical seq2seq LSTM | |
| Medaiyese et al. | Wavelet transform analytics for RF-based UAV detection and identification system using machine learning | |
| Wei et al. | Self-attention Bi-LSTM networks for radar signal modulation recognition | |
| CN111783558A (en) | Satellite navigation interference signal type intelligent identification method and system | |
| Valdenegro-Toro | Object recognition in forward-looking sonar images with convolutional neural networks | |
| CN113156391B (en) | Radar signal multi-dimensional feature intelligent sorting method | |
| Ozturk et al. | RF-based low-SNR classification of UAVs using convolutional neural networks | |
| Oh et al. | A UAV classification system based on FMCW radar micro-Doppler signature analysis | |
| CN113033473B (en) | ST2DCNN + SE-based radar overlapping signal identification method | |
| Ghadimi et al. | Deep learning-based approach for low probability of intercept radar signal detection and classification | |
| CN113156376B (en) | SACNN-based radar radiation source signal identification method | |
| Zhang et al. | A novel ship target detection algorithm based on error self-adjustment extreme learning machine and cascade classifier | |
| Yar et al. | A complete framework of radar pulse detection and modulation classification for cognitive EW | |
| CN115114949A (en) | Intelligent ship target identification method and system based on underwater acoustic signals | |
| Ristea et al. | Estimating the magnitude and phase of automotive radar signals under multiple interference sources with fully convolutional networks | |
| Jiang et al. | Continuous IFF response signal recognition technology based on capsule network | |
| CN114580461A (en) | Training method of radar interference recognition network and radar interference recognition method | |
| CN116797796A (en) | Signal identification method based on time-frequency analysis and deep learning under DRFM intermittent sampling | |
| Ibrahim et al. | A multimodel deep learning algorithm to detect North Atlantic right whale up-calls | |
| Zeng et al. | User-definable dynamic hand gesture recognition based on Doppler radar and few-shot learning | |
| CN117233706B (en) | Radar active interference identification method based on multilayer channel attention mechanism | |
| CN117289218B (en) | Active interference identification method based on attention cascade network | |
| CN114282576A (en) | Radar signal modulation format identification method and device based on time-frequency analysis and denoising | |
| Lu et al. | Sign language recognition with CW radar and machine learning | |
| CN117331031A (en) | LPI radar signal spectrogram fusion identification method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |