CN116540190A - End-to-end self-supervision intelligent interference suppression method and device and electronic equipment - Google Patents
End-to-end self-supervision intelligent interference suppression method and device and electronic equipment Download PDFInfo
- Publication number
- CN116540190A CN116540190A CN202310824849.XA CN202310824849A CN116540190A CN 116540190 A CN116540190 A CN 116540190A CN 202310824849 A CN202310824849 A CN 202310824849A CN 116540190 A CN116540190 A CN 116540190A
- Authority
- CN
- China
- Prior art keywords
- interference
- interference suppression
- network
- data
- radar echo
- 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
Links
- 230000001629 suppression Effects 0.000 title claims abstract description 126
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000001228 spectrum Methods 0.000 claims abstract description 56
- 238000012545 processing Methods 0.000 claims abstract description 48
- 238000010586 diagram Methods 0.000 claims abstract description 21
- 238000001514 detection method Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 53
- 238000003062 neural network model Methods 0.000 claims description 42
- 230000006870 function Effects 0.000 claims description 26
- 239000013598 vector Substances 0.000 claims description 22
- 238000004891 communication Methods 0.000 claims description 14
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 230000014509 gene expression Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 230000000116 mitigating effect Effects 0.000 claims 3
- 238000013528 artificial neural network Methods 0.000 abstract description 5
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000005764 inhibitory process Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000006467 substitution reaction Methods 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
- G01S7/023—Interference mitigation, e.g. reducing or avoiding non-intentional interference with other HF-transmitters, base station transmitters for mobile communication or other radar systems, e.g. using electro-magnetic interference [EMI] reduction techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computer Networks & Wireless Communication (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Mathematical Analysis (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses an end-to-end self-supervision intelligent interference suppression method, a device and electronic equipment, wherein the method comprises the following steps: acquiring first radar echo data to be processed, performing short-time Fourier transform, obtaining first time frequency spectrum data, and inputting the first time frequency spectrum data into an interference suppression network; the interference suppression network comprises a first sub-network and a second sub-network, wherein the first sub-network comprises a first convolution layer, a second convolution layer and a DropBlock layer; detecting whether interference exists or not by using a DropBlock layer according to a first feature diagram output by the second convolution layer, performing interference suppression when the interference exists, and outputting an interference suppression result diagram; and inputting the interference suppression result graph into a second sub-network to obtain first time spectrum data after anti-interference processing, and calculating first radar echo data after anti-interference processing. According to the invention, the interference detection, the suppression and the repair are finished through the end-to-end operation of the neural network once, so that the processing efficiency is improved.
Description
Technical Field
The invention belongs to the technical field of radars, and particularly relates to an end-to-end self-supervision intelligent interference suppression method and device and electronic equipment.
Background
Under the current complex and changeable electromagnetic interference environment background, the traditional interference suppression method has the problems of weak self-adaption capability, interference detection requirement and the like, so that the working performance of the radar is seriously affected. Currently, emerging artificial intelligence technologies such as deep learning and the like achieve excellent achievement in various fields based on a data driving principle, and an object of radar anti-interference processing is still data in nature, so that anti-interference research combined with an intelligent technology is a great trend of current development. However, the existing intelligent interference suppression method is mainly based on supervised learning to construct an interference suppression algorithm, but unstable interference data distribution in practical application can cause the algorithm to have a practical problem.
Specifically, in the prior art, a self-supervision learning thought is adopted, a neural network model is built based on a self-encoder, and a reconstruction error distribution of compression reconstruction is carried out on a time spectrum of a radar signal to be processed, so that detection and positioning of an interference time-frequency component in a time-frequency domain are realized, and a notch filter is built based on the detection and positioning to complete interference suppression; meanwhile, aiming at the signal loss problem caused by interference suppression, a neural network model is constructed to predict and repair the lost signal in the prior art, so that the intelligent interference suppression of the two-step method of interference suppression-signal repair is realized.
However, in the above method, in order to locate the interference in the echo signal in the time-frequency domain, a convolutional neural network model based on a self-encoder is constructed to compress and reconstruct the echo signal, so that the number of convolutional layers contained in the network model is large, resulting in large calculation amount, long processing time and low efficiency of the network model. In addition, in order to solve the problem of useful signal loss caused by the process of covering and suppressing interference, the prior art constructs a signal restoration neural network to predict and complement the lost signal, so that the processing process comprises two forward propagation operations of the neural network, and further the calculation complexity is increased.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides an end-to-end self-supervision intelligent interference suppression method, an end-to-end self-supervision intelligent interference suppression device and electronic equipment. The technical problems to be solved by the invention are realized by the following technical scheme:
in a first aspect, the present invention provides an end-to-end self-supervision intelligent interference suppression method, including:
acquiring first radar echo data to be processed;
performing short-time Fourier transform on the first radar echo data to obtain first time spectrum data;
inputting the first time spectrum data into an interference suppression network; wherein the interference suppression network is: the method comprises the steps that a neural network model is obtained through training based on a loss function constructed on an hypersphere in advance, the interference suppression network comprises a first sub-network and a second sub-network, and the first sub-network comprises a first convolution layer, a second convolution layer and a DropBlock layer which are sequentially connected;
detecting whether interference exists in the first radar echo data by utilizing the DropBlock layer according to the first characteristic diagram output by the second convolution layer, and performing interference suppression when the interference exists, and outputting an interference suppression result diagram;
inputting the interference suppression result graph into the second sub-network to obtain first time spectrum data after anti-interference processing;
and calculating first radar echo data after anti-interference processing according to the first time spectrum data after anti-interference processing.
In one embodiment of the invention, the interference suppression network is obtained by training the following steps:
acquiring undisturbed second radar echo data, and performing short-time Fourier transform on the second radar echo data to obtain a time spectrum data set;
respectively splicing the real part and the imaginary part of the second time spectrum data in the time spectrum data set to form a training data set containing a plurality of training data;
inputting a preset number of training data into a neural network model to be trained;
determining a loss value by using an output result of the neural network model to be trained, a second feature map output by a second convolution layer in the neural network model to be trained and a loss function, wherein the loss function is obtained based on 8-dimensional hypersphere construction;
updating network parameters of the neural network model to be trained and the radius of the hypersphere according to the loss value;
and when the preset iteration round number is reached, obtaining the interference suppression network with the training completed.
In one embodiment of the invention, the loss function is:
；
in the method, in the process of the invention,representing the entered training data +.>Training data representing said inputs +.>Output result processed by neural network model to be trained, < ->A second characteristic diagram representing the output of a second convolution layer in the neural network model to be trained,/->Representing +.>Second feature vector->Representing the number of second feature vectors in said second feature map, < >>Representing the mean square error loss function, ">Representing a structural similarity loss function, < >>Indicating transpose,/->Indicate->Updating the radius of the hypersphere obtained during round iteration, < >>Representing the current iteration round, +.>，Is the preset iteration round number.
In one embodiment of the invention, the mean square error loss functionAnd a structural similarity loss function->The respective expressions are as follows:
；
；
in the method, in the process of the invention,representing the output result +.>The%>Output data->Training data representing said inputs +.>The%>Training data->，/>Representing the output result +.>Middle->7X 7 first local data formed for the center point, ">Training data representing said inputs +.>Middle->Is the middle warmer7X 7 second local data formed by heart points, ">、/>Respectively indicate->And->Mean value of->、/>Respectively indicate->And->Variance of->Representation->And->Covariance of->、/>Are all preset constants.
In one embodiment of the present invention, the step of updating the network parameters of the neural network model to be trained and the radius of the hypersphere according to the loss value includes:
calculating gradients of loss values relative to network parameters of the neural network model to be trainedAnd a gradient +/with respect to the radius of the hypersphere>；
Using Adam optimization algorithm, gradientGradient->And preset learning rate->And updating network parameters of the neural network model to be trained and the radius of the hypersphere.
In one embodiment of the present invention, after the step of obtaining the trained interference suppression network, the method further comprises:
inputting training data input into a neural network model to be trained in the training process into the interference suppression network after training is completed, and obtaining a plurality of third feature graphs output by the second convolution layer;
and calculating the modulus value of each third feature vector in each third feature map, and taking the maximum value of the modulus values of the third feature vectors as the final radius of the hypersphere.
In one embodiment of the present invention, according to a first feature map output by the second convolution layer, the step of detecting whether there is interference in the first radar echo data by using the DropBlock layer, and performing interference suppression when there is interference, and outputting an interference suppression result map includes:
calculating a modulus value of each first feature vector in the first feature map output by the second convolution layer;
comparing the modulus value of the first feature vector with the final radius of the hypersphere;
and when the modulus value of the first eigenvector is larger than or equal to the final radius of the hypersphere, indicating that the interference exists in the first radar echo data, setting the first eigenvector with the modulus value larger than or equal to the final radius to 0, and outputting an interference suppression result graph.
In one embodiment of the present invention, the step of calculating the first radar echo data after the anti-interference processing according to the first time spectrum data after the anti-interference processing includes:
and performing inverse short-time Fourier transform on the first time spectrum data subjected to the anti-interference processing to obtain first radar echo data subjected to the anti-interference processing.
In a second aspect, the present invention provides an end-to-end self-supervised intelligent interference suppression apparatus, comprising:
the acquisition module is used for acquiring first radar echo data to be processed;
the processing module is used for carrying out short-time Fourier transform on the first radar echo data to obtain first time spectrum data;
a first input module for inputting the first time spectrum data into an interference suppression network; wherein the interference suppression network is: the method comprises the steps that a neural network model is obtained through training based on a loss function constructed on an hypersphere in advance, the interference suppression network comprises a first sub-network and a second sub-network, and the first sub-network comprises a first convolution layer, a second convolution layer and a DropBlock layer which are sequentially connected;
the detection module is used for detecting whether interference exists in the first radar echo data or not by utilizing the DropBlock layer according to the first characteristic diagram output by the second convolution layer, carrying out interference suppression when the interference exists, and outputting an interference suppression result diagram;
the second input module is used for inputting the interference suppression result graph into the second sub-network to obtain first time spectrum data after anti-interference processing;
the calculation module is used for calculating first radar echo data after the anti-interference processing according to the first time spectrum data after the anti-interference processing.
In a third aspect, the present invention further provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory complete communication with each other through the communication bus;
a memory for storing a computer program;
and a processor, configured to implement the method steps described in the first aspect when executing the program stored in the memory.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides an end-to-end self-supervision intelligent interference suppression method, an end-to-end self-supervision intelligent interference suppression device and electronic equipment, wherein an interference suppression network comprises a first sub-network and a second sub-network, wherein a shallow layer (namely the first sub-network) of the interference suppression network is used for detecting abnormal characteristics caused by interference in first radar echo data to be processed, so that the detection and suppression of interference are realized; further, after interference suppression, the deep layer (namely the second sub-network) of the interference suppression network is used for processing the lossy characteristics during interference suppression, so that the first radar echo data is repaired. Therefore, compared with the prior art, the invention can finish interference detection, inhibition and repair through one-time neural network operation end-to-end, thereby improving the processing efficiency.
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Drawings
Fig. 1 is a flowchart of an end-to-end self-supervision intelligent interference suppression method provided by an embodiment of the present invention;
fig. 2 is a schematic structural diagram of an interference suppression network according to an embodiment of the present invention;
fig. 3 is a flowchart of an interference suppression network training process provided by an embodiment of the present invention;
fig. 4 is a schematic structural diagram of an end-to-end self-supervision intelligent interference suppression device according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to specific examples, but embodiments of the present invention are not limited thereto.
Fig. 1 is a flowchart of an end-to-end self-supervision intelligent interference suppression method provided by an embodiment of the present invention, and fig. 2 is a schematic structural diagram of an interference suppression network provided by an embodiment of the present invention. As shown in fig. 1-2, an embodiment of the present invention provides an end-to-end self-supervision intelligent interference suppression method, which includes:
s11, acquiring first radar echo data to be processed;
s12, performing short-time Fourier transform on the first radar echo data to obtain first time spectrum data;
s13, inputting first time spectrum data into an interference suppression network; wherein, the interference suppression network is: the method comprises the steps that a neural network model is obtained through training based on a loss function constructed on an hypersphere in advance, an interference suppression network comprises a first sub-network and a second sub-network, and the first sub-network comprises a first convolution layer, a second convolution layer and a DropBlock layer which are sequentially connected;
s14, detecting whether interference exists in the first radar echo data by using a DropBlock layer according to the first feature map output by the second convolution layer, and performing interference suppression when the interference exists, and outputting an interference suppression result map;
s15, inputting an interference suppression result diagram into a second sub-network to obtain first time spectrum data after anti-interference processing;
s16, calculating first radar echo data after anti-interference processing according to the first time spectrum data after anti-interference processing.
In this embodiment, first radar echo data to be processed is obtained and short-time fourier transform is performed according to preset parameters, where the preset parameters include: the window function of the short-time Fourier transform is set to be a Hamming window, the window length is 31, the step length is 1, and the number of transformation points is 32, so that first time spectrum data is obtained. Because the first time spectrum data is complex, before the first time spectrum data is input into the interference suppression network, the real part and the imaginary part of the first time spectrum data can be respectively spliced in a channel dimension, and then the spliced real part and the spliced imaginary part are used as two channels to be input into the interference suppression network.
Optionally, the interference suppression network adopted in this embodiment is a neural network model obtained by training with a loss function constructed based on an hypersphere in advance, where a shallow layer of the interference suppression network, that is, a first subnetwork, determines whether interference exists in the first radar echo data by detecting abnormal features, if no interference exists in the first radar echo data, it is not necessary to perform subsequent processing on the first radar echo data, otherwise, if interference exists in the first radar echo data, it is necessary to suppress the interference, and repair is performed by using a deep layer of the interference suppression network, that is, a second subnetwork, so as to obtain the first radar echo data after the anti-interference processing.
It should be noted that, referring to fig. 2, in the interference suppression network, the first sub-network includes a first convolution layer, a second convolution layer and a DropBlock layer that are sequentially connected, and the second sub-network includes a third convolution layer, a fourth convolution layer, a first transpose convolution layer, a second transpose convolution layer, a third transpose convolution layer and a fourth transpose convolution layer that are sequentially connected.
Fig. 3 is a flowchart of an interference suppression network training process provided by an embodiment of the present invention. Optionally, as shown in fig. 3, the interference suppression network is obtained through training by the following steps:
s31, acquiring undisturbed second radar echo data, and performing short-time Fourier transform on the second radar echo data to obtain a time spectrum data set;
s32, respectively splicing the real part and the imaginary part of the second time spectrum data in the time spectrum data set to form a training data set containing a plurality of training data;
s33, inputting a preset number of training data into a neural network model to be trained, wherein the neural network model to be trained is a preset initial neural network model;
s34, determining a loss value by using an output result of the neural network model to be trained, a second feature map output by a second convolution layer in the neural network model to be trained and a loss function, wherein the loss function is obtained by constructing an 8-dimensional hypersphere;
s35, updating network parameters of the neural network model to be trained and the radius of the hypersphere according to the loss value;
and S36, obtaining the interference suppression network after training when the preset iteration round number is reached.
Specifically, in steps S31 to S32, 4000 undisturbed second radar echo data are selected, and then, according to preset parameters such as: performing short-time Fourier transform on the window function which is a Hamming window, the window length of 31, the step length of 1 and the number of transformation points of 32 to obtain a time spectrum data set containing a plurality of second time spectrum data; similarly, since the second time-frequency spectrum data is in a complex form, the real part and the imaginary part of each second time-frequency spectrum data in the time-frequency spectrum data set can be respectively subjected to channel dimension splicing to form a training data set.
Alternatively, the loss function used in the training process may be constructed from an 8-dimensional hypersphere with a constant sphere center as the origin and an initial radius value set to 0.3, and then the loss function is expressed as follows:
；
in the method, in the process of the invention,representing the entered training data +.>Training data +.>Output result processed by neural network model to be trained, < ->A second feature map representing the output of a second convolution layer in the neural network model to be trained, ++>Representing the +.f. in the second feature map>Second feature vector->Representing the number of second feature vectors in the second feature map,/-, for example>Representing the mean square error loss function, ">Representing a structural similarity loss function, < >>Indicating transpose,/->Indicate->Updating the radius of the obtained hypersphere during the iteration of the wheel, < >>Representing the current iteration round, +.>，/>Is the preset iteration round number.
Further, the mean square error loss functionAnd a structural similarity loss function->The respective expressions are as follows:
；
；
in the method, in the process of the invention,representing the output result +.>The%>Output data->Training data +.>The%>Training data->，/>Representing the output result +.>Middle->7X 7 first local data formed for the center point, ">Training data +.>Middle->7X 7 second local data formed for the center point,>、/>respectively indicate->And->Mean value of->、/>Respectively indicate->And->Variance of->Representation->And->Covariance of->、/>Are all preset constants.
In step S35, updating the network parameters of the neural network model to be trained and the radius of the hypersphere according to the loss value, including:
calculating gradients of loss values relative to network parameters of the neural network model to be trainedAnd a gradient with respect to the radius of the hypersphere +.>；
Using Adam optimization algorithm, gradientGradient->And preset learning rate->And updating network parameters of the neural network model to be trained and the radius of the hypersphere.
And (3) iteratively executing the steps S33-S35, and obtaining the interference suppression network after training when the iteration round number reaches the preset iteration round number. Optionally, the number of iteration rounds is presetK=10000, the preset learning rate is 0.0002.
Illustratively, the network parameters of the trained interference suppression model are specifically shown in the following table:
TABLE 1
In this embodiment, after the step of obtaining the trained interference suppression network, the method further includes:
inputting training data input into a neural network model to be trained in the training process into a trained interference suppression network to obtain a plurality of third feature images output by a second convolution layer;
and calculating the modulus value of each third feature vector in each third feature map, and taking the maximum value of the modulus values of the third feature vectors as the final radius of the hypersphere.
It should be noted that, the training is to make the feature of the training data be distributed in the minimum hypersphere, but the inventor finds that during the research process, the second feature vectors of the second feature map are all distributed in the hypersphere and cannot reach a steady state, so the radius of the hypersphere at the last iteration is not taken as the final radius, but the module value is calculated according to the distribution of the third feature vector of the training data, so that the interference detection and suppression results are more stable and accurate.
Optionally, in the step S14, according to the first feature map output by the second convolution layer, the step of detecting whether there is interference in the first radar echo data by using the DropBlock layer, performing interference suppression when there is interference, and outputting an interference suppression result map includes:
calculating the modulus value of each first feature vector in the first feature map output by the second convolution layer;
comparing the modulus value of the first feature vector with the final radius of the hypersphere;
when the modulus value of the first eigenvector is larger than or equal to the final radius of the hypersphere, which indicates that the interference exists in the first radar echo data, the first eigenvector with the modulus value larger than or equal to the final radius is set to 0, and an interference suppression result graph is output.
That is, the DropBlock layer processes the first feature vector according to the following formula:
；
in the method, in the process of the invention,represents the final radius of the hypersphere, < >>Representing the first feature in the first feature mapjA first one of the feature vectors is used to generate a first feature vector,。
it should be appreciated that the first feature vector in the first feature mapThe modulus of (2) corresponds to the first eigenvector +.>The distance between the first feature vector and the sphere center of the hypersphere is determined in the above manner, if the distance is greater than the final radius of the hypersphere, the interference suppression processing is performed, if the distance is less than the final radius of the hypersphere, no processing is performed, and the original data characteristics are reserved, namely, the DropBlock layer is utilized to realize the detection and suppression of interference.
Optionally, in step S31, the step of calculating the first radar echo data after the anti-interference processing according to the first time spectrum data after the anti-interference processing includes:
and performing inverse short-time Fourier transform on the first time spectrum data subjected to the anti-interference treatment to obtain first radar echo data subjected to the anti-interference treatment.
Fig. 4 is a schematic structural diagram of an end-to-end self-supervision intelligent interference suppression device according to an embodiment of the present invention. As shown in fig. 4, an embodiment of the present invention provides an end-to-end self-supervision intelligent interference suppression device, which includes:
an acquisition module 410, configured to acquire first radar echo data to be processed;
the processing module 420 is configured to perform short-time fourier transform on the first radar echo data to obtain first time spectrum data;
a first input module 430 for inputting first time-spectrum data into an interference suppression network; wherein, the interference suppression network is: the method comprises the steps that a neural network model is obtained through training based on a loss function constructed on an hypersphere in advance, an interference suppression network comprises a first sub-network and a second sub-network, and the first sub-network comprises a first convolution layer, a second convolution layer and a DropBlock layer which are sequentially connected;
the detection module 440 is configured to detect whether interference exists in the first radar echo data by using the DropBlock layer according to the first feature map output by the second convolution layer, and perform interference suppression when the interference exists, and output an interference suppression result map;
a second input module 450, configured to input the interference suppression result graph into a second sub-network, to obtain first time spectrum data after the anti-interference processing;
the calculating module 460 is configured to calculate the first radar echo data after the anti-interference processing according to the first time spectrum data after the anti-interference processing.
Fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. The embodiment of the invention also provides an electronic device, as shown in fig. 5, which comprises a processor 501, a communication interface 502, a memory 503 and a communication bus 504, wherein the processor 501, the communication interface 502 and the memory 503 complete communication with each other through the communication bus 504,
a memory 503 for storing a computer program;
the processor 501 is configured to execute the program stored in the memory 503, and implement the following steps:
acquiring first radar echo data to be processed;
performing short-time Fourier transform on the first radar echo data to obtain first time spectrum data;
inputting the first time spectrum data into an interference suppression network; wherein, the interference suppression network is: the method comprises the steps that a neural network model is obtained through training based on a loss function constructed on an hypersphere in advance, an interference suppression network comprises a first sub-network and a second sub-network, and the first sub-network comprises a first convolution layer, a second convolution layer and a DropBlock layer which are sequentially connected;
detecting whether interference exists in the first radar echo data by using a DropBlock layer according to a first feature diagram output by the second convolution layer, and performing interference suppression when the interference exists, and outputting an interference suppression result diagram;
inputting the interference suppression result graph into a second sub-network to obtain first time spectrum data after anti-interference processing;
and calculating first radar echo data after anti-interference processing according to the first time spectrum data after anti-interference processing.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides an end-to-end self-supervision intelligent interference suppression method, an end-to-end self-supervision intelligent interference suppression device and electronic equipment, wherein an interference suppression network comprises a first sub-network and a second sub-network, wherein a shallow layer (namely the first sub-network) of the interference suppression network is used for detecting abnormal characteristics caused by interference in first radar echo data to be processed, so that the detection and suppression of interference are realized; further, after interference suppression, the deep layer (namely the second sub-network) of the interference suppression network is used for processing the lossy characteristics during interference suppression, so that the first radar echo data is repaired. Therefore, compared with the prior art, the invention can finish interference detection, inhibition and repair through one-time neural network operation end-to-end, thereby improving the processing efficiency.
In the description of the present invention, the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. In the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.
The description of the terms "one embodiment," "some embodiments," "example," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Further, one skilled in the art can engage and combine the different embodiments or examples described in this specification.
Although the present application has been described herein in connection with various embodiments, other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed application, from a review of the figures, the disclosure, and the appended claims.
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 (10)
1. An end-to-end self-supervision intelligent interference suppression method is characterized by comprising the following steps:
acquiring first radar echo data to be processed;
performing short-time Fourier transform on the first radar echo data to obtain first time spectrum data;
inputting the first time spectrum data into an interference suppression network; wherein the interference suppression network is: the method comprises the steps that a neural network model is obtained through training based on a loss function constructed on an hypersphere in advance, the interference suppression network comprises a first sub-network and a second sub-network, and the first sub-network comprises a first convolution layer, a second convolution layer and a DropBlock layer which are sequentially connected;
detecting whether interference exists in the first radar echo data by utilizing the DropBlock layer according to the first characteristic diagram output by the second convolution layer, and performing interference suppression when the interference exists, and outputting an interference suppression result diagram;
inputting the interference suppression result graph into the second sub-network to obtain first time spectrum data after anti-interference processing;
and calculating first radar echo data after anti-interference processing according to the first time spectrum data after anti-interference processing.
2. The end-to-end self-supervised intelligent interference suppression method according to claim 1, wherein the interference suppression network is trained to:
acquiring undisturbed second radar echo data, and performing short-time Fourier transform on the second radar echo data to obtain a time spectrum data set;
respectively splicing the real part and the imaginary part of the second time spectrum data in the time spectrum data set to form a training data set containing a plurality of training data;
inputting a preset number of training data into a neural network model to be trained;
determining a loss value by using an output result of the neural network model to be trained, a second feature map output by a second convolution layer in the neural network model to be trained and a loss function, wherein the loss function is obtained based on 8-dimensional hypersphere construction;
updating network parameters of the neural network model to be trained and the radius of the hypersphere according to the loss value;
and when the preset iteration round number is reached, obtaining the interference suppression network with the training completed.
3. The end-to-end self-supervised intelligent interference mitigation method of claim 2, wherein the loss function is:
；
in the method, in the process of the invention,representing the entered training data +.>Training data representing said inputs +.>Output result processed by neural network model to be trained, < ->A second characteristic diagram representing the output of a second convolution layer in the neural network model to be trained,/->Representing +.>Second feature vector->Representing the number of second feature vectors in said second feature map, < >>Representing the mean square error loss function, ">Representing a structural similarity loss function, < >>Indicating transpose,/->Represent the firstkUpdating the resulting radius of the hypersphere at 1 iteration,krepresenting the current iteration round, +.>，/>Is the preset iteration round number.
4. The end-to-end self-supervised intelligent interference mitigation method of claim 3, wherein the mean square error loss functionAnd a structural similarity loss function->The respective expressions are as follows:
；
；
in the method, in the process of the invention,representing the output result +.>The%>Output data->Training data representing said inputs +.>The%>Training data->，/>Representing the output result +.>Middle->7X 7 first local data formed for the center point, ">Training data representing said inputs +.>Middle->7 x 7 second local data formed for the center point,、/>respectively indicate->And->Mean value of->、/>Respectively indicate->And->Variance of->Representation->And->Covariance of->、/>Are all preset constants.
5. The end-to-end self-supervised intelligent interference mitigation method of claim 3, wherein updating network parameters of the neural network model to be trained and the radius of the hypersphere according to the loss value comprises:
calculating gradients of loss values relative to network parameters of the neural network model to be trainedAnd a gradient +/with respect to the radius of the hypersphere>；
Using Adam optimization algorithm, gradientGradient->And preset learning rate->And updating network parameters of the neural network model to be trained and the radius of the hypersphere.
6. The end-to-end self-supervised intelligent interference suppression method according to claim 2, further comprising, after the step of obtaining the trained interference suppression network:
inputting training data input into a neural network model to be trained in the training process into the interference suppression network after training is completed, and obtaining a plurality of third feature graphs output by the second convolution layer;
and calculating the modulus value of each third feature vector in each third feature map, and taking the maximum value of the modulus values of the third feature vectors as the final radius of the hypersphere.
7. The end-to-end self-supervision intelligent interference suppression method according to claim 6, wherein the step of detecting whether interference exists in the first radar echo data by using the DropBlock layer according to the first feature map output by the second convolution layer, performing interference suppression when interference exists, and outputting an interference suppression result map includes:
calculating a modulus value of each first feature vector in the first feature map output by the second convolution layer;
comparing the modulus value of the first feature vector with the final radius of the hypersphere;
and when the modulus value of the first eigenvector is larger than or equal to the final radius of the hypersphere, indicating that the interference exists in the first radar echo data, setting the first eigenvector with the modulus value larger than or equal to the final radius to 0, and outputting an interference suppression result graph.
8. The end-to-end self-supervised intelligent interference suppression method according to claim 1, wherein the step of calculating the first radar echo data after the anti-interference processing according to the first time spectrum data after the anti-interference processing comprises:
and performing inverse short-time Fourier transform on the first time spectrum data subjected to the anti-interference processing to obtain first radar echo data subjected to the anti-interference processing.
9. An end-to-end self-supervising intelligent interference suppression device, comprising:
the acquisition module is used for acquiring first radar echo data to be processed;
the processing module is used for carrying out short-time Fourier transform on the first radar echo data to obtain first time spectrum data;
a first input module for inputting the first time spectrum data into an interference suppression network; wherein the interference suppression network is: the method comprises the steps that a neural network model is obtained through training based on a loss function constructed on an hypersphere in advance, the interference suppression network comprises a first sub-network and a second sub-network, and the first sub-network comprises a first convolution layer, a second convolution layer and a DropBlock layer which are sequentially connected;
the detection module is used for detecting whether interference exists in the first radar echo data or not by utilizing the DropBlock layer according to the first characteristic diagram output by the second convolution layer, carrying out interference suppression when the interference exists, and outputting an interference suppression result diagram;
the second input module is used for inputting the interference suppression result graph into the second sub-network to obtain first time spectrum data after anti-interference processing;
the calculation module is used for calculating first radar echo data after the anti-interference processing according to the first time spectrum data after the anti-interference processing.
10. The electronic equipment is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;
a memory for storing a computer program;
a processor for carrying out the method steps of any one of claims 1-8 when executing a program stored on a memory.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310824849.XA CN116540190A (en) | 2023-07-06 | 2023-07-06 | End-to-end self-supervision intelligent interference suppression method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310824849.XA CN116540190A (en) | 2023-07-06 | 2023-07-06 | End-to-end self-supervision intelligent interference suppression method and device and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116540190A true CN116540190A (en) | 2023-08-04 |
Family
ID=87449238
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310824849.XA Pending CN116540190A (en) | 2023-07-06 | 2023-07-06 | End-to-end self-supervision intelligent interference suppression method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116540190A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200341109A1 (en) * | 2019-03-14 | 2020-10-29 | Infineon Technologies Ag | Fmcw radar with interference signal suppression using artificial neural network |
| CN113791389A (en) * | 2021-05-18 | 2021-12-14 | 西安电子科技大学 | Unknown radar radiation source individual identification method and system |
| CN114494846A (en) * | 2021-12-18 | 2022-05-13 | 南京理工大学 | Hyperspectral image classification method based on multi-scale internal volume network |
| CN114779185A (en) * | 2022-04-29 | 2022-07-22 | 西安电子科技大学 | Radar signal anti-interference low-loss recovery method of coding and decoding convolutional neural network |
| US20220291329A1 (en) * | 2019-11-28 | 2022-09-15 | HELLA GmbH & Co. KGaA | Method for identifying interference in a radar system |
| CN115097391A (en) * | 2022-06-20 | 2022-09-23 | 东南大学 | Synthetic aperture radar interference suppression method based on automatic encoder |
| CN116299219A (en) * | 2023-05-18 | 2023-06-23 | 西安电子科技大学 | Interference depth characteristic distance measurement combined detection and suppression method |
-
2023
- 2023-07-06 CN CN202310824849.XA patent/CN116540190A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200341109A1 (en) * | 2019-03-14 | 2020-10-29 | Infineon Technologies Ag | Fmcw radar with interference signal suppression using artificial neural network |
| US20220291329A1 (en) * | 2019-11-28 | 2022-09-15 | HELLA GmbH & Co. KGaA | Method for identifying interference in a radar system |
| CN113791389A (en) * | 2021-05-18 | 2021-12-14 | 西安电子科技大学 | Unknown radar radiation source individual identification method and system |
| CN114494846A (en) * | 2021-12-18 | 2022-05-13 | 南京理工大学 | Hyperspectral image classification method based on multi-scale internal volume network |
| CN114779185A (en) * | 2022-04-29 | 2022-07-22 | 西安电子科技大学 | Radar signal anti-interference low-loss recovery method of coding and decoding convolutional neural network |
| CN115097391A (en) * | 2022-06-20 | 2022-09-23 | 东南大学 | Synthetic aperture radar interference suppression method based on automatic encoder |
| CN116299219A (en) * | 2023-05-18 | 2023-06-23 | 西安电子科技大学 | Interference depth characteristic distance measurement combined detection and suppression method |
Non-Patent Citations (1)
| Title |
|---|
| 韩朝赟等: "纹理异常感知SAR自监督学习干扰抑制方法", 雷达学报, vol. 12, no. 1 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110728360A (en) | Micro-energy device energy identification method based on BP neural network | |
| CN111650453B (en) | Power equipment diagnosis method and system based on windowing characteristic Hilbert imaging | |
| CN112269134A (en) | Battery SOC and SOH joint estimation method based on deep learning | |
| CN111985383A (en) | Transient electromagnetic signal noise separation and identification method based on improved variational modal decomposition | |
| CN111797974A (en) | Power system state estimation method combining particle filtering and convolutional neural network | |
| CN115656817A (en) | Permanent magnet synchronous motor turn-to-turn short circuit fault detection method based on neural network technology | |
| CN112861066A (en) | Machine learning and FFT (fast Fourier transform) -based blind source separation information source number parallel estimation method | |
| CN115563444A (en) | Signal reconstruction method and device, computer equipment and storage medium | |
| CN117290788B (en) | Power distribution network fault identification method and system based on improved wavelet transformation algorithm | |
| CN116299219B (en) | Interference depth characteristic distance measurement combined detection and suppression method | |
| CN114429151A (en) | Magnetotelluric signal identification and reconstruction method and system based on depth residual error network | |
| CN111461303B (en) | Digital core reconstruction method and system based on generation of antagonistic neural network | |
| CN116540190A (en) | End-to-end self-supervision intelligent interference suppression method and device and electronic equipment | |
| CN117192358A (en) | Permanent magnet synchronous motor turn-to-turn short circuit fault diagnosis method based on convolutional neural network | |
| CN116754213A (en) | Electric gate valve fault diagnosis method, device and equipment based on strong noise background | |
| CN114978586B (en) | Power grid attack detection method and system based on attack genes and electronic equipment | |
| CN115906303A (en) | Planar microwave filter design method and device based on machine learning | |
| CN112016684B (en) | Electric power terminal fingerprint identification method of deep parallel flexible transmission network | |
| CN116975742A (en) | Partial discharge pattern recognition method, apparatus, device, and storage medium | |
| CN117892117B (en) | Fault positioning method and system for power transmission line of power distribution network | |
| CN116738820A (en) | Electromagnetic response optimization system, method and product for passive device | |
| CN116112932B (en) | Data knowledge dual-drive radio frequency fingerprint identification method and system | |
| CN112612602B (en) | Automatic compression processing method for target detection network model | |
| CN117572300A (en) | Motor turn-to-turn short circuit fault detection method based on variational modal decomposition fusion deep learning | |
| CN117932230B (en) | Fault diagnosis method based on characteristic decoupling and sample recombination under small sample condition |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230804 |