CN107315996A - A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA - Google Patents
A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA Download PDFInfo
- Publication number
- CN107315996A CN107315996A CN201710359078.6A CN201710359078A CN107315996A CN 107315996 A CN107315996 A CN 107315996A CN 201710359078 A CN201710359078 A CN 201710359078A CN 107315996 A CN107315996 A CN 107315996A
- Authority
- CN
- China
- Prior art keywords
- imf
- eigenmatrix
- ships
- pca
- feature
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA, extracts ships noise signal average, variance, peak value, degree of bias time domain charactreristic parameter, constitutes temporal signatures matrix；Using EEMD methods, former ships noise signal is decomposed, IMF components are obtained, and IMF components are converted into the feature parameter vectors, so as to observe the change from frequency band energy feature, constituted frequency domain character vector, eigenmatrix A is collectively formed with temporal signatures vector；Dimension-reduction treatment is done to eigenmatrix A using PCA dimension reduction methods, high-dimensional eigenmatrix is mapped to low dimensional as new feature, constitutive characteristic matrix B；Train LSSVM parameters, according to principle of minimization risk, adjusting parameter γ and core width cs, finally the eigenmatrix B vectors after dimensionality reduction are input in grader, classification results are test, the inventive method can combine time-domain and frequency-domain information, form message complementary sense, more complete characteristic information is provided, and nicety of grading can be improved.
Description
Technical field
The present invention relates to Underwater Targets Recognition field, especially a kind of ships noise under water based on IMF Energy-Entropies and PCA
Feature extracting method.
Background technology
At present, China is continuously increased to the demand of marine resources, and development and utilization scale is continued to increase.Underwater Targets Recognition
Technology is also developed rapidly, and has been applied to all many-sides, such as fish detection, marine organisms research and the detection of underwater submarine
Deng.Underwater Targets Recognition is for biological protection and strengthens ocean defence safety and plays an important role, for ocean exploitation and
Sustainable development is significant.And the identification of ships noise signal under water and classification are the emphasis of Underwater Targets Recognition classification
And difficult point, the classification of ships is mainly judged by the engine noise of ships.In complicated marine environment, the underwater sound such as ships
The mechanism of production of echo signal radiated noise is sufficiently complex, and composition is various.Simultaneously because the complicated and changeable and water of underwater acoustic channel
The multi-path effect of acoustic signal propagation, makes underwater sound signal that non-gaussian, non-stationary, nonlinear " three is non-" property is often presented.And pass
The method of system is all based on signal and noise is steady and Gaussian random process this hypothesis, with Acoustic Object vibration and noise reducing
The raising of performance, is difficult the accurate feature for extracting underwater radiation noise based on the Fourier classical signal processing methods converted.
And the characteristic vector data amount that traditional method for extracting is arrived is big, easily causes dimension calamity, easily causes over-fitting to show in grader
As reducing nicety of grading.Therefore complete characteristic information is extracted, reduction data dimension turns into underwater sound signal feature extraction and classification
Key.
The content of the invention
Present invention aims at provide a kind of accurate extraction noise signal feature, reduction operand, improve nicety of grading
The noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA.
To achieve the above object, following technical scheme is employed：The method of the invention comprises the following steps：
Step 1, ships noise signal average, variance, peak value, degree of bias time domain charactreristic parameter are extracted, as temporal signatures to
Amount, constitutes temporal signatures matrix；
Step 2, using the improved EEMD methods of EMD, former ships noise signal is decomposed, IMF components are obtained, and by IMF
Component is converted into the feature parameter vectors, so as to observe the change from frequency band energy feature, frequency domain character vector is constituted, with step 1
In temporal signatures vector collectively form eigenmatrix A；
Step 3, dimension-reduction treatment is done to eigenmatrix A using PCA dimension reduction methods, high-dimensional eigenmatrix be mapped to low
Dimension represents former feature, constitutive characteristic matrix B as far as possible as new feature；
Step 4, LSSVM (least square method supporting vector machine) parameter is trained, according to principle of minimization risk, adjusting parameter γ
With core width cs, finally the eigenmatrix B vectors after dimensionality reduction are input in grader, test classification results, and with non-dimensionality reduction
Eigenmatrix classification results are contrasted.
Further, in step 2, the EMD methods are empirical mode decomposition, can be by the intrinsic mode of Dynamic Signal point
Amount (IMF) is extracted；The EEMD methods are to give signal addition minimum amplitude white noise on the basis of EMD methods, are utilized
The characteristics of white noise spectrum equiblibrium mass distribution, carrys out the interruptive area of equalizing signal, so as to remove modal aliasing；The EEMD methods are carried
The Energy-Entropy of the IMF components taken constitutes frequency domain character matrix, and constitutes ships noise characteristic with the temporal signatures vector in step 1
Matrix A, extracts more complete, comprehensive feature.
Further, in step 3, the time and frequency domain characteristics matrix A extracted is combined with PCA, dimension-reduction treatment is done, reduced
Data dimension, it is to avoid over-fitting occurs in grader.
Compared with prior art, the inventive method has the following advantages that：
1st, propose the method that is combined of time-domain and frequency-domain information, using average, variance, peak value, the temporal signatures such as the degree of bias and
The eigenmatrix that the frequency domain character that the IMF Energy-Entropies tried to achieve are constituted is collectively constituted so that the characteristic information of acquisition is more complete, more entirely
Face；
2nd, it is too high for the eigenmatrix dimension of traditional method for extracting, dimension calamity is easily produced, so as to be produced in grader
Raw over-fitting problem, using PCA dimension reduction methods, the principal component and contribution rate of the former eigenmatrix of analysis, by contribution rate of accumulative total>
95% principal component replaces original all evaluation indexes, reaches dimensionality reduction purpose.
Brief description of the drawings
Fig. 1 is the flow chart of the inventive method.
Fig. 2 is the flow chart of IMF Energy-Entropies in the inventive method.
Fig. 3 is the sampled signal figure of sample.
Embodiment
The present invention will be further described below in conjunction with the accompanying drawings：
As shown in figure 1, the method for the invention comprises the following steps：
Step 1, the time domain charactreristic parameters such as ships noise signal average, variance, peak value, the degree of bias are extracted, temporal signatures are used as
Vector, constitutes temporal signatures matrix；
Step 2, EEMD decomposition is carried out to former ships noise signal, empirical mode decomposition (EMD) is field of signal processing solution
The steady, new method of nonlinear properties by no means, it can extract the intrinsic modal components (IMF) of Dynamic Signal, but
In practical application, due to end effect or intermittent influence so that EMD, which is decomposed, causes modal aliasing phenomenon.And integrated experience
Mode Decomposition (EEMD) method can overcome this phenomenon, and it adds minimum amplitude white noise to signal, utilizes white noise spectrum
The characteristics of equiblibrium mass distribution, carrys out the interruptive area of equalizing signal, so as to remove modal aliasing.IMF components are obtained after decomposition, it is contemplated that
The different frequency composition of unlike signal, and IMF amplitude differences, frequency domain character is constituted simultaneously using IMF Energy-Entropy as characteristic vector
Be combined with time domain charactreristic parameter, form message complementary sense, so as to further improve discrimination, with the temporal signatures in step 1 to
Amount collectively forms eigenmatrix A；
Step 3, dimension-reduction treatment is done to eigenmatrix A using PCA dimension reduction methods, high-dimensional eigenmatrix be mapped to low
Dimension represents former feature, constitutive characteristic matrix B as far as possible as new feature；Generally, contribution rate of accumulative total is chosen
>90% principal component replaces all principal components, and regard the product of the corresponding characteristic vector of characteristic value and original matrix as new spy
Input matrix is levied into grader；
Step 4, LSSVM (least square method supporting vector machine) parameter is trained, according to principle of minimization risk, adjusting parameter γ
With core width cs, finally the eigenmatrix B vectors after dimensionality reduction are input in grader, test classification results, and with non-dimensionality reduction
Eigenmatrix classification results are contrasted.
Embodiment 1：
The inventive method includes following steps：
(1) temporal signatures information is extracted
The time domain charactreristic parameter of ships noise signal is extracted, as temporal signatures vector, its parameter is respectively：
Average
A=mean (x)
Variance
Peak value
F=max (x)
The degree of bias
Wherein, xiIt is sampled data,For average, E is desired value, and root mean square RMS formula is：
T=[A, D, B, F], average, variance, peak value, degree of bias information reflects that the time domain of vibration signal is special from different perspectives
Levy, and with certain complementarity, this four time domain parameters are combined, form four-dimension vector T=[A, D, B, F], can
Using the temporal signatures vector as ships noise.
(2) frequency domain information IMF Energy-Entropies are extracted
With EMD improved model EEMD and comentropy method, the frequency domain character of ships noise signal is extracted.Such as accompanying drawing 2
It is shown, first, in the original vibration signal x (t) superposition average be zero, standard deviation be constant random Gaussian white noise wi(t),
Obtain signal x to be decomposedi(t)
xi(t)=x (t)+wi(t)
Wherein, wi(t) be ith add white Gaussian noise after signal.
EMD decomposition is carried out to above formula, j-th of IMF components f is can obtain after circulating j timesij(t),
Circulate after n times, calculate IMF population mean：
Energy-Entropy concept is introduced on this basis, seeks the ENERGY E of each IMF componentsi
The energy of each IMF components is normalized：
pm=Em/E
Seek each IMF Energy-Entropy：
IMF energy entropy is distinguished to the distribution situation between different mode as frequency domain character vector.
(3) PCA dimensionality reductions
Now take A, B, C, the class ships noises of D tetra- include training sample 30 per noise like, test sample 30, then at present
Eigenvectors matrix be 120*16, using PCA dimension-reduction treatment, it is standardized first, dimension and quantity is eliminated
The limitation of level, the calculation formula of standardization is as follows：
In formula, u representation vectors X mean vector；Std (X) represents that X standard vector is poor；For the result after standard.
Then decorrelation coefficient matrix R, correlation matrix has reacted the size of the linear correlation degree of 2 variables.Phase
Relation number is bigger, shows that the linear relationship between 2 variables is closer；Coefficient correlation is smaller, indicates the linear correlation between 2 variables
Degree is smaller.
In formula, rij(ij=1,2 ..., it is p) former variable xiWith xjCoefficient correlation, its calculation formula is：
WithThe average value of respectively i-th and j-th index；N and p are respectively sample number and data dimension.
Eigen vector is calculated, det (R- λ I)=0 is made, obtains its characteristic value, and it is sized,
λ1≥λ2≥…≥λp.Correspondence and characteristic root λ are finally obtained respectivelyiCharacteristic vector ui。
Calculate contribution rate and contribution rate of accumulative total, k-th of principal component ykVariance contribution ratio be
Principal component ymContribution rate of accumulative total be
Selection contribution rate of accumulative total reaches 95% preceding i principal component, now think i principal component instead of original n
Item index.Obtain after principal component, the characteristic vector before being extracted according to the descending order of characteristic value corresponding to i characteristic value
μi(1,2 ..., the n) projection matrix constituted.Initial data is multiplied with projection matrix, the score of principal component is finally given, as
Characteristic vector is input in grader.
(4) classified using least square method supporting vector machine
Sorter model uses the SVM improved models LSSVM of the propositions such as SUYUKENS.Assuming that there are m class classification samples, then instruct
Practice sample and be set to S1：{(xi,yi), i=1,2 ... n, wherein, yiIt is the m dimensional vectors of 1 and -1 composition, when sample belongs to jth,
Then yiJth be 1, other are -1.Characteristic vector battle array is now set to X, classification battle array is set to Y, then their the i-th row is respectively xi
And yi。
SVM problem with inequality constraint is converted into problem with equality constraint by LSSVM：
In formula, w and b are Optimal Separating Hyperplane,Relevant parameter,For classification empiric risk,Represent compartment away from numerical values recited, γ ＞ 0 are penalty factor, in training balanced learning machine answer
Polygamy and empiric risk.
To solve above-mentioned optimization problem, Lagrangian is introduced：
For the variable w in Lagrange multiplier, b, e, α seeks extreme value, transformed rear available：
In formula, yiFor Y the i-th row, Ωjk (t)=yij.yik.k(xj,xk), unit matrix I ∈ RN×N, αtAnd biRespectively correspond to
yiLagrange multiplier vector sum constant terms, LS-SVM functions can be set up：
In formula, x is test data, xiFor training data；K () is kernel function, generally selects RBF type functions：
(xi,xj)=exp (- | | xi-xj||2)/σ2
Finally, by discriminant function Cj(x)=sign (y,j(x)) (j=1,2 ... m) determine x status categories.
The recognition effect of the embodiment of the present invention can be further illustrated by following emulation data：
(1) A, B, C are now taken, tetra- kinds of samples of D take each 30 of training sample respectively, each 30 of test sample, A, B, C, D's
Noise signal sample graph is as shown in Figure 3.
The frequency domain information such as the time-domain information of training sample and IMF Energy-Entropies is extracted respectively, 120*16 feature is constituted
Matrix, as shown in table 1：
The sample time-domain of table 1 and IMF energy entropy
After PCA dimensionality reductions, principal component contributor rate is chosen>95% principal component, and by its corresponding characteristic value and former square
The product of battle array is input in grader as new eigenmatrix：Such as table 2
Eigenmatrix (preceding 4*10) after table 2PCA dimensionality reductions
Enter data into grader and tested, not dimensionality reduction data classification is contrasted, as a result as follows：
The discrimination of table 3 is contrasted
As shown in Table 3, before discrimination of the discrimination that IMF comentropies are combined with time-domain information after dimensionality reduction is than non-dimensionality reduction
Discrimination improve 5% or so, it was demonstrated that the validity of the inventive method.
Embodiment described above is only that the preferred embodiment of the present invention is described, not to the model of the present invention
Enclose and be defined, on the premise of design spirit of the present invention is not departed from, technical side of the those of ordinary skill in the art to the present invention
In various modifications and improvement that case is made, the protection domain that claims of the present invention determination all should be fallen into.
Claims (3)
1. a kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA, it is characterised in that methods described bag
Include following steps：
Step 1, ships noise signal average, variance, peak value, degree of bias time domain charactreristic parameter are extracted, temporal signatures vector, structure is used as
Into temporal signatures matrix；
Step 2, using the improved EEMD methods of EMD, former ships noise signal is decomposed, IMF components are obtained, and by IMF components
The feature parameter vectors are converted into, so as to observe the change from frequency band energy feature, frequency domain character vector are constituted, and in step 1
Temporal signatures vector collectively forms eigenmatrix A；
Step 3, dimension-reduction treatment is done to eigenmatrix A using PCA dimension reduction methods, high-dimensional eigenmatrix is mapped to low dimensional
As new feature, and former feature, constitutive characteristic matrix B are represented as far as possible；
Step 4, LSSVM (least square method supporting vector machine) parameter, according to principle of minimization risk, adjusting parameter γ and core are trained
Eigenmatrix B vectors after dimensionality reduction, are finally input in grader by width cs, test classification results, and with the feature of non-dimensionality reduction
Matrix Classification result is contrasted.
2. a kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA according to claim 1, its
It is characterised by：In step 2, the EMD methods are empirical mode decomposition, can be by the intrinsic modal components (IMF) of Dynamic Signal
Extract；The EEMD methods are to give signal addition minimum amplitude white noise on the basis of EMD methods, utilize white noise audio frequency
The characteristics of spectral balancing is distributed carrys out the interruptive area of equalizing signal, so as to remove modal aliasing；IMF points that the EEMD methods are extracted
The Energy-Entropy of amount constitutes frequency domain character matrix, and constitutes ships noise characteristic matrix A with the temporal signatures vector in step 1, carries
Take more complete, comprehensive feature.
3. a kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA according to claim 1, its
It is characterised by：In step 3, the time and frequency domain characteristics matrix A extracted is combined with PCA, dimension-reduction treatment is done, data dimension is reduced
Degree, it is to avoid over-fitting occurs in grader.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710359078.6A CN107315996A (en) | 2017-05-19 | 2017-05-19 | A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710359078.6A CN107315996A (en) | 2017-05-19 | 2017-05-19 | A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107315996A true CN107315996A (en) | 2017-11-03 |
Family
ID=60181495
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710359078.6A Pending CN107315996A (en) | 2017-05-19 | 2017-05-19 | A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107315996A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108506736A (en) * | 2018-03-19 | 2018-09-07 | 浙江理工大学 | A kind of invasion signal diagnostic system and method |
| CN108509850A (en) * | 2018-02-24 | 2018-09-07 | 华南理工大学 | A kind of invasion signal Recognition Algorithm based on distribution type fiber-optic system |
| CN109634151A (en) * | 2018-12-27 | 2019-04-16 | 中国舰船研究设计中心 | A kind of method of adjust automatically ship acoustics state |
| CN109784410A (en) * | 2019-01-18 | 2019-05-21 | 西安邮电大学 | A kind of feature extraction and classification method of ships radiated noise signal |
| CN110161125A (en) * | 2019-06-17 | 2019-08-23 | 哈尔滨工业大学 | The Aeroengine Smart monitoring method combined based on acceleration with sound emission cognition technology |
| CN112183280A (en) * | 2020-09-21 | 2021-01-05 | 西安交通大学 | Underwater sound target radiation noise classification method and system based on EMD and compressed sensing |
| CN112633333A (en) * | 2020-12-11 | 2021-04-09 | 广州致新电力科技有限公司 | Method for identifying partial discharge defects |
| CN112818830A (en) * | 2021-01-29 | 2021-05-18 | 上海跃磁生物科技有限公司 | Magnetocardiogram signal noise reduction method, system, medium and device |
| CN113552447A (en) * | 2021-07-27 | 2021-10-26 | 上海电机学院 | Series arc fault detection method based on random forest |
| CN114462458A (en) * | 2022-04-11 | 2022-05-10 | 自然资源部第一海洋研究所 | Ship underwater signal noise reduction and target enhancement method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103226132A (en) * | 2013-04-25 | 2013-07-31 | 哈尔滨工业大学 | High speed railway steel rail flaw detection experiment platform and detection method |
| CN104916059A (en) * | 2015-05-18 | 2015-09-16 | 中国民航大学 | EEMD-based M-Z interferometer type airport perimeter early warning system pattern recognition method |
| US20150293143A1 (en) * | 2013-07-21 | 2015-10-15 | National Chiao Tung University | Feature Extraction from Human Gaiting Patterns using Principal Component Analysis and Multivariate Empirical Mode Decomposition |
| CN105136454A (en) * | 2015-10-15 | 2015-12-09 | 上海电机学院 | Wind turbine gear box fault recognition method |
| CN105628425A (en) * | 2016-01-01 | 2016-06-01 | 三峡大学 | Rotation machinery early stage fault diagnosis method based on heredity annealing optimization multi-core support vector machine |
| CN106529428A (en) * | 2016-10-31 | 2017-03-22 | 西北工业大学 | Underwater target recognition method based on deep learning |
-
2017
- 2017-05-19 CN CN201710359078.6A patent/CN107315996A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103226132A (en) * | 2013-04-25 | 2013-07-31 | 哈尔滨工业大学 | High speed railway steel rail flaw detection experiment platform and detection method |
| US20150293143A1 (en) * | 2013-07-21 | 2015-10-15 | National Chiao Tung University | Feature Extraction from Human Gaiting Patterns using Principal Component Analysis and Multivariate Empirical Mode Decomposition |
| CN104916059A (en) * | 2015-05-18 | 2015-09-16 | 中国民航大学 | EEMD-based M-Z interferometer type airport perimeter early warning system pattern recognition method |
| CN105136454A (en) * | 2015-10-15 | 2015-12-09 | 上海电机学院 | Wind turbine gear box fault recognition method |
| CN105628425A (en) * | 2016-01-01 | 2016-06-01 | 三峡大学 | Rotation machinery early stage fault diagnosis method based on heredity annealing optimization multi-core support vector machine |
| CN106529428A (en) * | 2016-10-31 | 2017-03-22 | 西北工业大学 | Underwater target recognition method based on deep learning |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108509850B (en) * | 2018-02-24 | 2022-03-29 | 华南理工大学 | Intrusion signal identification method based on distributed optical fiber system |
| CN108509850A (en) * | 2018-02-24 | 2018-09-07 | 华南理工大学 | A kind of invasion signal Recognition Algorithm based on distribution type fiber-optic system |
| CN108506736B (en) * | 2018-03-19 | 2019-11-29 | 浙江理工大学 | A kind of invasion signal diagnostic system and method |
| CN108506736A (en) * | 2018-03-19 | 2018-09-07 | 浙江理工大学 | A kind of invasion signal diagnostic system and method |
| CN109634151A (en) * | 2018-12-27 | 2019-04-16 | 中国舰船研究设计中心 | A kind of method of adjust automatically ship acoustics state |
| CN109634151B (en) * | 2018-12-27 | 2024-03-22 | 中国舰船研究设计中心 | Method for automatically adjusting acoustic state of ship |
| CN109784410A (en) * | 2019-01-18 | 2019-05-21 | 西安邮电大学 | A kind of feature extraction and classification method of ships radiated noise signal |
| CN109784410B (en) * | 2019-01-18 | 2022-11-22 | 西安邮电大学 | Characteristic extraction and classification method for ship radiation noise signals |
| CN110161125A (en) * | 2019-06-17 | 2019-08-23 | 哈尔滨工业大学 | The Aeroengine Smart monitoring method combined based on acceleration with sound emission cognition technology |
| CN112183280A (en) * | 2020-09-21 | 2021-01-05 | 西安交通大学 | Underwater sound target radiation noise classification method and system based on EMD and compressed sensing |
| CN112633333A (en) * | 2020-12-11 | 2021-04-09 | 广州致新电力科技有限公司 | Method for identifying partial discharge defects |
| CN112818830A (en) * | 2021-01-29 | 2021-05-18 | 上海跃磁生物科技有限公司 | Magnetocardiogram signal noise reduction method, system, medium and device |
| CN112818830B (en) * | 2021-01-29 | 2022-11-25 | 上海跃磁生物科技有限公司 | Magnetocardiogram signal noise reduction method, system, medium and device |
| CN113552447A (en) * | 2021-07-27 | 2021-10-26 | 上海电机学院 | Series arc fault detection method based on random forest |
| CN114462458A (en) * | 2022-04-11 | 2022-05-10 | 自然资源部第一海洋研究所 | Ship underwater signal noise reduction and target enhancement method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107315996A (en) | A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA | |
| Qu et al. | Radar signal intra-pulse modulation recognition based on convolutional denoising autoencoder and deep convolutional neural network | |
| CN109389058A (en) | Sea clutter and noise signal classification method and system | |
| Liu et al. | Deep learning and recognition of radar jamming based on CNN | |
| CN112364779A (en) | Underwater sound target identification method based on signal processing and deep-shallow network multi-model fusion | |
| CN106529428A (en) | Underwater target recognition method based on deep learning | |
| CN105261367B (en) | A kind of method for distinguishing speek person | |
| CN105550636A (en) | Method and device for identifying target types | |
| CN113571067B (en) | Voiceprint recognition countermeasure sample generation method based on boundary attack | |
| CN109637545A (en) | Based on one-dimensional convolution asymmetric double to the method for recognizing sound-groove of long memory network in short-term | |
| CN108922515A (en) | Speech model training method, audio recognition method, device, equipment and medium | |
| CN108091345B (en) | Double-ear voice separation method based on support vector machine | |
| Yang et al. | Open set recognition of underwater acoustic targets based on GRU-CAE collaborative deep learning network | |
| CN114897023B (en) | Underwater sound target identification method based on extraction of sensitive difference features of underwater sound targets | |
| Cao et al. | Underwater target classification at greater depths using deep neural network with joint multiple‐domain feature | |
| CN113111786B (en) | Underwater target identification method based on small sample training diagram convolutional network | |
| Sun et al. | Radar emitter individual identification based on convolutional neural network learning | |
| CN107403618A (en) | Based on the audio event sorting technique and computer equipment for stacking base rarefaction representation | |
| CN114372487A (en) | Underwater sound target identification method based on multi-discriminant generation countermeasure network | |
| Zhang et al. | A highly stealthy adaptive decay attack against speaker recognition | |
| CN112052880A (en) | Underwater sound target identification method based on weight updating support vector machine | |
| Yang et al. | CNN-GRUA-FC stock price forecast model based on multi-factor analysis | |
| CN106709428A (en) | One-dimensional distance image stable identification method based on Euler kernel principle component analysis (KPCA) | |
| CN116430317A (en) | Radiation source modulation pattern and individual identification method and system | |
| Xie et al. | Data augmentation and deep neural network classification based on ship radiated noise |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20171103 |
|
| WD01 | Invention patent application deemed withdrawn after publication |