CN111325143A - Underwater target identification method under unbalanced data set condition - Google Patents

Underwater target identification method under unbalanced data set condition Download PDF

Info

Publication number
CN111325143A
CN111325143A CN202010100399.6A CN202010100399A CN111325143A CN 111325143 A CN111325143 A CN 111325143A CN 202010100399 A CN202010100399 A CN 202010100399A CN 111325143 A CN111325143 A CN 111325143A
Authority
CN
China
Prior art keywords
target
frequency characteristic
time
neural network
training set
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010100399.6A
Other languages
Chinese (zh)
Inventor
姜喆
董亚芬
申晓红
王海燕
闫永胜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN202010100399.6A priority Critical patent/CN111325143A/en
Publication of CN111325143A publication Critical patent/CN111325143A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)

Abstract

The invention provides an underwater target recognition method under the condition of unbalanced data set, and provides a cost-sensitive convolutional neural network aiming at the problem that the existing underwater target recognition method cannot accurately recognize a few types of samples in unbalanced data set.

Description

Underwater target identification method under unbalanced data set condition
Technical Field
The invention belongs to the field of information signal processing, relates to the fields of underwater signal processing, neural networks and the like, and particularly relates to an underwater target identification method.
Background
The underwater target recognition is one of key technologies for realizing ocean intelligent perception, and has important significance for maintaining ocean safety and building ocean strong countries. The traditional underwater target identification method mainly comprises the steps of artificially extracting the characteristics of target radiation noise according to domain knowledge, then constructing classifiers such as a support vector machine and the like, and carrying out classification and identification on targets based on the extracted characteristics. In recent years, deep learning methods typified by convolutional neural networks have been greatly developed in the fields of image recognition, voice recognition, and the like. Experts and scholars also begin to apply deep learning methods such as convolutional neural networks to the field of underwater target identification, and compared with the traditional method, the method improves the accuracy rate of underwater target identification and makes a breakthrough progress.
In an actual marine environment, the frequency degree of activities of different types of underwater targets is often different, so that the data volume of different types of underwater targets is obviously different. The imbalance of the data quantity can cause the characteristics of the minority class not to be fully expressed, and the classifier can hardly learn the distinguishing boundary between the majority class and the minority class accurately, so that the classification effect is not ideal, and great difficulty is brought to the identification of the minority class target. The existing underwater target identification method is difficult to accurately identify a few samples in the unbalanced data set, and the identification effect of the unbalanced data set is not ideal. Therefore, a method suitable for underwater target identification under unbalanced data set conditions is needed.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides an underwater target identification method under the condition of unbalanced data set, and the method adds influence factors to the traditional cross entropy loss function by utilizing a cost sensitive convolutional neural network according to the sample prediction probability, thereby realizing the purpose of self-adaptively adjusting the loss according to the sample prediction probability. The more accurate the sample prediction is, the smaller the value of the influence factor added to the cross entropy loss function is, and the smaller the loss is; and vice versa. The cost-sensitive convolutional neural network provided by the invention can improve the identification accuracy of a few samples on the premise of ensuring the identification accuracy of a plurality of samples in an unbalanced data set.
The technical scheme adopted by the invention for solving the technical problem comprises the following steps:
step 1, distributing hydrophones in the sea, and collecting and recording radiation noise of a first target and a second target which are different, wherein the data volume of the two targets is unbalanced;
step 2, performing short-time Fourier transform on the collected target radiation noise to obtain time-frequency characteristic graphs of two types of target radiation noise signals; carrying out size adjustment and pixel value normalization processing on the time-frequency characteristic diagram, and labeling the processed time-frequency characteristic diagram for distinguishing the time-frequency characteristic diagrams of two types of targets;
step 3, after the time-frequency characteristic graphs of the two types of targets obtained in the step 2 are respectively randomly disturbed, dividing the time-frequency characteristic graphs according to a set proportion to respectively form a target one training set, a target one testing set, a target two training set and a target two testing set; the target one training set and the target two training set form an overall training set, and the target one testing set and the target two testing set form an overall testing set;
and 4, constructing a cost sensitive convolutional neural network, and comprising the following steps:
step 401, constructing a convolutional neural network having m convolutional layers, m pooling layers, and n fully-connected layers, where m and n are any positive integer;
step 402, defining a loss function
Figure BDA0002386707120000021
Wherein the true class of the sample is class c, pcIs the probability that the neural network prediction sample belongs to class c; log pcIs the cross entropy loss;
Figure BDA0002386707120000022
is to predict the probability p from the samplescInfluence factors added for cross entropy loss, wherein the value of the influence factors is between 0 and 1, β is a hyperparameter which takes a non-negative value, pcThe value range of (1) is between 0 and 1;
step 5, training the convolutional neural network constructed in the step 401 on the overall training set in the step 3, and optimizing by adopting an Adam algorithm by taking the loss function defined in the step 402 as a target function to obtain an optimal network model parameter;
and 6, testing the optimal model obtained in the step 5 on the overall test set in the step 3, and giving the identification accuracy of the first target and the second target in the overall test set.
The proportion of the training set and the test set divided in the step 3 is 7: 3 or 6: 4 or 8: 2.
the invention has the beneficial effects that: aiming at the problem that the existing underwater target identification method cannot accurately identify the few samples in the unbalanced data set, the cost-sensitive convolutional neural network is provided, and the identification accuracy of the few samples can be improved on the premise of ensuring the identification accuracy of the many samples in the unbalanced data set. The cost sensitive convolutional neural network provided by the present invention can achieve this effect because the loss function of the network defined in step 402 can open the gap between the loss of the correct prediction sample and the loss of the incorrect prediction sample, reduce the loss of the correct prediction sample in a relative sense, and increase the loss of the incorrect prediction sample. In particular, p for correctly predicting samplescThe value must be greater than p of the mispredicted samplecValue, then, cross entropy loss of correctly predicted sample-log pcAnd influencing factors
Figure BDA0002386707120000023
Must be less than the cross-entropy loss of the mispredicted sample
Figure BDA0002386707120000026
And influencing factors
Figure BDA0002386707120000024
And-logpcNot less than 0 and
Figure BDA0002386707120000025
this is always true for all samples. Thus, the loss of correctly predicted samples
Figure BDA0002386707120000031
Must be less than the loss of mispredicted samples
Figure BDA0002386707120000032
Adjusting the value of the over-parameter β affects the gap between the loss of correctly predicted samples and the loss of incorrectly predicted samples.
Drawings
FIG. 1 is a general method flow diagram of the present invention.
Fig. 2 is a diagram of a convolutional neural network structure in an embodiment of the present invention.
Detailed Description
Aiming at the problem that the existing underwater target identification method cannot accurately identify a few samples in an unbalanced data set, the invention discloses a cost-sensitive convolutional neural network, which mainly comprises the following steps:
step 1, distributing hydrophones in the sea, and collecting and recording radiation noise of a first target and a second target which are different, wherein the data volume of the two targets is unbalanced;
and 2, performing short-time Fourier transform on the collected target radiation noise to obtain time-frequency characteristic graphs of two types of target radiation noise signals. Carrying out size adjustment and pixel value normalization processing on the time-frequency characteristic diagram, labeling the processed time-frequency characteristic diagram, wherein the labels of the time-frequency characteristic diagrams of the two types of targets are 1 and 0 respectively, and forming a data set;
and 3, after the time-frequency characteristic graphs of the two types of targets obtained in the step 2 are respectively randomly disturbed, dividing the time-frequency characteristic graphs according to a certain proportion to respectively form a target one training set, a target one testing set, a target two training set and a target two testing set, wherein the division proportion can be 7: 3 or 6: 4 or 8: 2, etc.; the target one training set and the target two training set form an overall training set, and the target one testing set and the target two testing set form an overall testing set;
and 4, constructing a cost sensitive convolutional neural network, and comprising the following steps:
step 401, constructing a convolutional neural network having m convolutional layers, m pooling layers, and n fully-connected layers, where m and n are any positive integer;
step 402, defining a loss function
Figure BDA0002386707120000033
Wherein the true class of the sample is class c, pcIs the probability that the neural network prediction sample belongs to class c; log pcIs the cross entropy loss;
Figure BDA0002386707120000034
is to predict the probability p from the samplescInfluence factors added for cross entropy loss, wherein the value of the influence factors is between 0 and 1, β is a hyperparameter which takes a non-negative value, pcThe value range of (1) is between 0 and 1; p is a radical ofcLarger, meaning more accurate network predictions, then the impact factor added for cross-entropy loss
Figure BDA0002386707120000035
The smaller the value of (a), the smaller the loss; and vice versa; the property of the exponential function enables the loss function to pull apart the difference between the loss of the correct prediction sample and the loss of the wrong prediction sample, so that the loss of the correct prediction sample is reduced in a relative sense, and the loss of the wrong prediction sample is improved, so that the network has cost sensitivity;
step 5, training the convolutional neural network constructed in the step 401 on the overall training set in the step 3, and optimizing by adopting an Adam algorithm by taking the loss function defined in the step 402 as a target function to obtain an optimal network model parameter;
and 6, testing the optimal model obtained in the step 5 on the overall test set in the step 3, and giving the identification accuracy of the first target and the second target in the overall test set.
The present invention will be further described with reference to the following drawings and examples, which include, but are not limited to, the following examples.
The invention discloses a cost-sensitive convolutional neural network, which aims at solving the problem that the existing underwater target identification method cannot accurately identify a few samples in an unbalanced data set. The network adds an influence factor to the traditional cross entropy loss function according to the sample prediction probability, thereby realizing the purpose of self-adaptively adjusting the loss according to the sample prediction probability. The more accurate the sample prediction is, the smaller the value of the influence factor added to the cross entropy loss function is, and the smaller the loss is; and vice versa. The cost-sensitive convolutional neural network provided by the invention can improve the identification accuracy of a few samples on the premise of ensuring the identification accuracy of a plurality of samples in an unbalanced data set. As shown in fig. 1, the present invention comprises the steps of:
step 1, distributing hydrophones in the sea, and collecting and recording radiation noise of a first target and a second target which are different, wherein the data volume of the two targets is unbalanced;
carrying out size adjustment and pixel value normalization processing on the time-frequency characteristic graph, wherein the size of the time-frequency characteristic graph can be adjusted to 64 × 64, labeling the processed time-frequency characteristic graph, and the labels of the time-frequency characteristic graphs of the two types of targets are 1 and 0 respectively to form a data set;
and 3, after the time-frequency characteristic graphs of the two types of targets obtained in the step 2 are disordered respectively at random, dividing the time-frequency characteristic graphs according to a certain proportion to respectively form a target one training set, a target one testing set, a target two training set and a target two testing set, wherein the division proportion can be 7: 3; the target one training set and the target two training set form an overall training set, and the target one testing set and the target two testing set form an overall testing set;
and 4, constructing a cost sensitive convolutional neural network, and comprising the following steps:
step 401, constructing a convolutional neural network having m convolutional layers, m pooling layers and n fully-connected layers, where m and n are any positive integer, m can be 3, and n can be 1, and the structure of the constructed convolutional neural network is shown in fig. 2;
step 402, defining a loss function
Figure BDA0002386707120000041
Wherein the true class of the sample is class c, pcIs the probability that the neural network prediction sample belongs to class c; log pcIs the cross entropy loss;
Figure BDA0002386707120000042
is to predict the probability p from the samplescInfluence factors added for cross entropy loss, wherein the value of the influence factors is between 0 and 1, β is a hyperparameter, takes a non-negative value, β can take values of 1, 2 and 3, and p iscThe value range of (1) is between 0 and 1; p is a radical ofcLarger, meaning more accurate network predictions, then the impact factor added for cross-entropy loss
Figure BDA0002386707120000051
The smaller the value of (a), the smaller the loss; and vice versa; the property of the exponential function enables the loss function to pull apart the difference between the loss of the correct prediction sample and the loss of the wrong prediction sample, so that the loss of the correct prediction sample is reduced in a relative sense, and the loss of the wrong prediction sample is improved, so that the network has cost sensitivity;
step 5, training the convolutional neural network constructed in the step 401 on the overall training set in the step 3, and optimizing by adopting an Adam algorithm by taking the loss function defined in the step 402 as a target function to obtain an optimal network model parameter;
and 6, testing the optimal model obtained in the step 5 on the overall test set in the step 3, and giving the identification accuracy of the first target and the second target in the overall test set.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.
It should be particularly noted that, for clarity of description, the underwater object identification method under unbalanced data set conditions provided by the present invention is illustrated in the technical solution, the drawings and the detailed description by taking only two types of objects as examples. The method is also suitable for identifying the underwater multi-class target data volume under the condition of unbalance.

Claims (2)

1. An underwater target identification method under the condition of unbalanced data set is characterized by comprising the following steps:
step 1, distributing hydrophones in the sea, and collecting and recording radiation noise of a first target and a second target which are different, wherein the data volume of the two targets is unbalanced;
step 2, performing short-time Fourier transform on the collected target radiation noise to obtain time-frequency characteristic graphs of two types of target radiation noise signals; carrying out size adjustment and pixel value normalization processing on the time-frequency characteristic diagram, and labeling the processed time-frequency characteristic diagram for distinguishing the time-frequency characteristic diagrams of two types of targets;
step 3, after the time-frequency characteristic graphs of the two types of targets obtained in the step 2 are respectively randomly disturbed, dividing the time-frequency characteristic graphs according to a set proportion to respectively form a target one training set, a target one testing set, a target two training set and a target two testing set; the target one training set and the target two training set form an overall training set, and the target one testing set and the target two testing set form an overall testing set;
and 4, constructing a cost sensitive convolutional neural network, and comprising the following steps:
step 401, constructing a convolutional neural network having m convolutional layers, m pooling layers, and n fully-connected layers, where m and n are any positive integer;
step 402, defining a loss function
Figure FDA0002386707110000011
Wherein the true class of the sample is class c, pcIs the probability that the neural network prediction sample belongs to class c; -logpcIs the cross entropy loss;
Figure FDA0002386707110000012
is to predict the probability p from the samplescInfluence factors added for cross entropy loss, wherein the value of the influence factors is between 0 and 1, β is a hyperparameter which takes a non-negative value, pcThe value range of (1) is between 0 and 1;
step 5, training the convolutional neural network constructed in the step 401 on the overall training set in the step 3, and optimizing by adopting an Adam algorithm by taking the loss function defined in the step 402 as a target function to obtain an optimal network model parameter;
and 6, testing the optimal model obtained in the step 5 on the overall test set in the step 3, and giving the identification accuracy of the first target and the second target in the overall test set.
2. The method for underwater object identification under data set imbalance conditions according to claim 1, wherein: the proportion of the training set and the test set divided in the step 3 is 7: 3 or 6: 4 or 8: 2.
CN202010100399.6A 2020-02-18 2020-02-18 Underwater target identification method under unbalanced data set condition Pending CN111325143A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010100399.6A CN111325143A (en) 2020-02-18 2020-02-18 Underwater target identification method under unbalanced data set condition

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010100399.6A CN111325143A (en) 2020-02-18 2020-02-18 Underwater target identification method under unbalanced data set condition

Publications (1)

Publication Number Publication Date
CN111325143A true CN111325143A (en) 2020-06-23

Family

ID=71168896

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010100399.6A Pending CN111325143A (en) 2020-02-18 2020-02-18 Underwater target identification method under unbalanced data set condition

Country Status (1)

Country Link
CN (1) CN111325143A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112232144A (en) * 2020-09-27 2021-01-15 西北工业大学 Personnel overboard detection and identification method based on improved residual error neural network
CN112396807A (en) * 2020-10-27 2021-02-23 西北工业大学 Method for detecting and identifying person falling into water
CN112528775A (en) * 2020-11-28 2021-03-19 西北工业大学 Underwater target classification method

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106952338A (en) * 2017-03-14 2017-07-14 网易（杭州）网络有限公司 Method, system and the readable storage medium storing program for executing of three-dimensional reconstruction based on deep learning
CN107454598A (en) * 2017-07-20 2017-12-08 浙江大学 One kind emulates attack detection method based on the unbalanced primary users of I/Q
CN107516312A (en) * 2017-08-14 2017-12-26 北京工业大学 A kind of Chinese medicine complexion automatic classification method using shallow-layer neutral net
CN108009525A (en) * 2017-12-25 2018-05-08 北京航空航天大学 A kind of specific objective recognition methods over the ground of the unmanned plane based on convolutional neural networks
CN108846380A (en) * 2018-04-09 2018-11-20 北京理工大学 A kind of facial expression recognizing method based on cost-sensitive convolutional neural networks
CN108846323A (en) * 2018-05-28 2018-11-20 哈尔滨工程大学 A kind of convolutional neural networks optimization method towards Underwater Targets Recognition
CN109325504A (en) * 2018-09-07 2019-02-12 中国农业大学 A kind of underwater sea cucumber recognition methods and system

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106952338A (en) * 2017-03-14 2017-07-14 网易（杭州）网络有限公司 Method, system and the readable storage medium storing program for executing of three-dimensional reconstruction based on deep learning
CN107454598A (en) * 2017-07-20 2017-12-08 浙江大学 One kind emulates attack detection method based on the unbalanced primary users of I/Q
CN107516312A (en) * 2017-08-14 2017-12-26 北京工业大学 A kind of Chinese medicine complexion automatic classification method using shallow-layer neutral net
CN108009525A (en) * 2017-12-25 2018-05-08 北京航空航天大学 A kind of specific objective recognition methods over the ground of the unmanned plane based on convolutional neural networks
CN108846380A (en) * 2018-04-09 2018-11-20 北京理工大学 A kind of facial expression recognizing method based on cost-sensitive convolutional neural networks
CN108846323A (en) * 2018-05-28 2018-11-20 哈尔滨工程大学 A kind of convolutional neural networks optimization method towards Underwater Targets Recognition
CN109325504A (en) * 2018-09-07 2019-02-12 中国农业大学 A kind of underwater sea cucumber recognition methods and system

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
ABDELOUAHID BEN TAMOU等: "Transfer Learning with deep Convolutional Neural Network for Underwater Live Fish Recognition", 《2018 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING, APPLICATIONS AND SYSTEMS (IPAS)》 *
GUOPING XU等: "A novel exponential loss function for pathological lymph node image classification", 《MIPPR 2019》 *
SALMAN H. KHAN等: "Cost-Sensitive Learning of Deep Feature Representations From Imbalanced Data", 《IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 》 *
刘有用等: "不平衡数据集下的水下目标快速识别方法", 《计算机工程与应用》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112232144A (en) * 2020-09-27 2021-01-15 西北工业大学 Personnel overboard detection and identification method based on improved residual error neural network
CN112396807A (en) * 2020-10-27 2021-02-23 西北工业大学 Method for detecting and identifying person falling into water
CN112528775A (en) * 2020-11-28 2021-03-19 西北工业大学 Underwater target classification method

Similar Documents

Publication Publication Date Title
CN110767216B (en) Voice recognition attack defense method based on PSO algorithm
CN109102005B (en) Small sample deep learning method based on shallow model knowledge migration
CN111325143A (en) Underwater target identification method under unbalanced data set condition
CN108600135A (en) A kind of recognition methods of signal modulation mode
CN102760444B (en) Support vector machine based classification method of base-band time-domain voice-frequency signal
CN109357749A (en) A kind of power equipment audio signal analysis method based on DNN algorithm
CN108922560A (en) A kind of city noise recognition methods based on interacting depth neural network model
CN113571067B (en) Voiceprint recognition countermeasure sample generation method based on boundary attack
CN110930995B (en) Voice recognition model applied to power industry
CN109581339B (en) Sonar identification method based on automatic adjustment self-coding network of brainstorming storm
CN103941131A (en) Transformer fault detecting method based on simplified set unbalanced SVM (support vector machine)
CN114337880B (en) Spectrum sensing method considering random arrival and departure of main user signals
CN109884591A (en) A kind of multi-rotor unmanned aerial vehicle acoustical signal Enhancement Method based on microphone array
CN103994820B (en) A kind of moving target recognition methods based on micropore diameter microphone array
CN116866129A (en) Wireless communication signal detection method
CN114143040A (en) Confrontation signal detection method based on multi-channel feature reconstruction
Kong et al. Radar waveform recognition using Fourier-based synchrosqueezing transform and CNN
CN109002810A (en) Model evaluation method, Radar Signal Recognition method and corresponding intrument
CN112347844A (en) Signal countermeasure sample detector design method based on LID
CN113109782B (en) Classification method directly applied to radar radiation source amplitude sequence
CN104882140A (en) Voice recognition method and system based on blind signal extraction algorithm
Yu et al. Histogram transform model using MFCC features for text-independent speaker identification
CN111259861A (en) Underwater acoustic communication signal modulation mode identification method under data set unbalanced condition
CN108269566B (en) Rifling wave identification method based on multi-scale sub-band energy set characteristics
CN114519372B (en) One-dimensional range profile target recognition method based on support vector machine

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

Application publication date: 20200623

RJ01 Rejection of invention patent application after publication