CN109740635A - Classification and Identification Feature Mapping method based on two classifiers - Google Patents
Classification and Identification Feature Mapping method based on two classifiers Download PDFInfo
- Publication number
- CN109740635A CN109740635A CN201811512493.1A CN201811512493A CN109740635A CN 109740635 A CN109740635 A CN 109740635A CN 201811512493 A CN201811512493 A CN 201811512493A CN 109740635 A CN109740635 A CN 109740635A
- Authority
- CN
- China
- Prior art keywords
- msf
- classification
- characteristic
- feature
- dimensional
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000013507 mapping Methods 0.000 title claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000005315 distribution function Methods 0.000 claims description 8
- 102100035373 Cyclin-D-binding Myb-like transcription factor 1 Human genes 0.000 claims description 7
- 101000804518 Homo sapiens Cyclin-D-binding Myb-like transcription factor 1 Proteins 0.000 claims description 7
- 230000008859 change Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000000926 separation method Methods 0.000 description 3
- 241000219098 Parthenocissus Species 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 244000287680 Garcinia dulcis Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Landscapes
- Geophysics And Detection Of Objects (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The Classification and Identification Feature Mapping method based on two classifiers that the present invention relates to a kind of, comprising the following steps: count the Gaussian Profile situation for training the every one-dimensional characteristic of all kinds of target signatures of classifier coefficient respectively；The Gaussian Profile situation of all kinds of every one-dimensional characteristics of target signature in the case where being counted respectively relative to all kinds of target signatures generation characteristic offset for training classifier coefficient；Distribution map relationship is constructed respectively to every one-dimensional characteristic in two classifier each in the case of characteristic offset both front and back according to the Gaussian Profile situation that statistics obtains.The present invention can be avoided the case where causing Classification and Identification accuracy rate to be greatly lowered because of characteristic offset.
Description
Technical field
The present invention relates to digital signal processing technique fields, special more particularly to a kind of Classification and Identification based on two classifiers
Levy mapping method.
Background technique
Classification and Identification technology is a kind of important Digital Signal Processing, is obtained in digital signal processing technique field
It is widely applied.And feature extraction and optimization are the first steps of Classification and Identification technology and are a vital steps, feature choosing
Select it is appropriate whether directly influence the design complexities of entire identifying system, and determine the identification accuracy of system.It is common
Feature mainly include temporal signatures, frequency domain character, parameterized model feature, high-order spectrum signature, time and frequency domain characteristics, non-linear spy
Sign etc., especially time and frequency domain characteristics play an increasingly important role in nonstationary random response, also achieve much at
Fruit.But since the non-stationary signal in actual signal treatment process is influenced by several factors, the signal of same target type
Signal characteristic under difficult environmental conditions is also widely different, such as Micro-seismic Signals feature can be with the variation of geological conditions
And change, voice signal property is affected by wind very big.It is therefore more likely that there is designed classifying and identifying system in certain class ring
Classification and Identification effect is fine under the conditions of border, and shifts to performance under in addition a kind of environmental condition and then sharply decline.
Summary of the invention
The Classification and Identification Feature Mapping method based on two classifiers that technical problem to be solved by the invention is to provide a kind of,
Avoid the case where causing Classification and Identification accuracy rate to be greatly lowered because of characteristic offset.
The technical solution adopted by the present invention to solve the technical problems is: providing a kind of Classification and Identification based on two classifiers
Feature Mapping method, comprising the following steps:
(1) the Gaussian Profile situation for training the every one-dimensional characteristic of all kinds of target signatures of classifier coefficient is counted respectively；
(2) all kinds of in the case where counting respectively relative to all kinds of target signatures generation characteristic offset for training classifier coefficient
The Gaussian Profile situation of the every one-dimensional characteristic of target signature；
(3) the Gaussian Profile situation obtained according to statistics is to every in two classifier each in the case of characteristic offset both front and back
One-dimensional characteristic constructs distribution map relationship respectively.
The step (1) and step (2) are all made of the Gaussian Profile situation of following method statistic: defining a kind of target signature
Number is N, intrinsic dimensionality M, characteristic sequence F1,...,FN(Fi=[Fi1,...,FiM]), then kth dimensional feature sequence is
F1k,...,FNk, calculate the mean value of kth dimensional feature sequenceCalculate the standard deviation of kth dimensional feature sequence
The step (3) specifically includes following sub-step:
(31) it defines in two classifiers for training the mean value of two class target kth dimensional feature sequences of classifier coefficient
For MTFk1And MTFk2, standard deviation STDTFk1And STDTFk2, two class target kth dimensional feature sequences in the case where characteristic offset occur
The mean value of column is MSFk1And MSFk2, standard deviation STDSFk1And STDSFk2；
(32) defining x-axis overturning parameter is DP=(MTFk1-MTFk2)×(MSFk1-MSFk2), as x-axis overturns parameter DP <
0, then x-axis is overturn, i.e. x=-x, MSFk1=-MSFk1, MSFk2=-MSFk2, otherwise it is directly entered step (33)；
(33) gauss of distribution function of two class target kth dimensional feature sequences is constructed according to the mean value and standard deviation that are calculated
(34) gauss of distribution function DF is calculated1(x) and DF2(x) intersection point xC；
(35) defining distribution map regional determination parameter is DMTF=sign (MTFk1-MTFk2), as distribution map region is sentenced
Determine parameter DMTF × (x-xC) > 0, kth dimensional feature distribution map relationship isIt is no
Then kth dimensional feature distribution map relationship is
If the number of hits being calculated is greater than 1 in the step (34), intersection point is chosen in MSFk1、MSFk2Between x
It is worth the separation as x-axis.
Beneficial effect
Due to the adoption of the above technical solution, compared with prior art, the present invention having the following advantages that and actively imitating
Fruit: the present invention is united respectively by the Gaussian Profile situation to every one-dimensional characteristic in the case of generation characteristic offset both front and back
Then meter constructs distribution map relationship to every one-dimensional characteristic in each two classifier according to the distribution situation that statistics obtains respectively,
The feature that characteristic offset occurs is corrected by the distribution map relationship of building.This method characteristic offset can occurring
Situation effectively corrects Classification and Identification feature, so as to avoid causing Classification and Identification accuracy rate to be greatly lowered because of characteristic offset
The case where occur.
Detailed description of the invention
Fig. 1 is flow chart of the invention.
Specific embodiment
Present invention will be further explained below with reference to specific examples.It should be understood that these embodiments are merely to illustrate the present invention
Rather than it limits the scope of the invention.In addition, it should also be understood that, after reading the content taught by the present invention, those skilled in the art
Member can make various changes or modifications the present invention, and such equivalent forms equally fall within the application the appended claims and limited
Range.
Embodiments of the present invention are related to a kind of Classification and Identification Feature Mapping method based on two classifiers, including following step
It is rapid: to count the Gaussian Profile situation for training the every one-dimensional characteristic of all kinds of target signatures of classifier coefficient respectively；It counts respectively
All kinds of target signatures are every in the case that characteristic offset (non-time-varying) occurs for all kinds of target signatures relative to training classifier coefficient
The Gaussian Profile situation of one-dimensional characteristic；The distribution situation obtained according to statistics is to each two points in the case of characteristic offset both front and back
Every one-dimensional characteristic constructs distribution map relationship respectively in class device.
Wherein, when counting Gaussian Profile situation with the following method: defining a kind of target signature number is N, intrinsic dimensionality
For M, characteristic sequence F1,...,FN(Fi=[Fi1,...,FiM]), then kth dimensional feature sequence is F1k,...,FNk, calculate kth dimension
The mean value of characteristic sequenceCalculate the standard deviation of kth dimensional feature sequence
When constructing distribution map relationship, first define in two classifiers for training two class targets of classifier coefficient
The mean value of kth dimensional feature sequence is MTFk1And MTFk2, standard deviation STDTFk1And STDTFk2, in the case where characteristic offset occurs
The mean value of two class target kth dimensional feature sequences is MSFk1And MSFk2, standard deviation STDSFk1And STDSFk2, and according to above-mentioned side
Method is calculated.Then, defining x-axis overturning parameter is DP=(MTFk1-MTFk2)×(MSFk1-MSFk2), as x-axis overturns parameter
DP < 0, then overturn x-axis, i.e. x=-x, MSFk1=-MSFk1, MSFk2=-MSFk2；Then, according to the mean value that is calculated with
Standard deviation constructs the gauss of distribution function of two class target kth dimensional feature sequencesAnd it counts
Calculate gauss of distribution function DF1(x) and DF2(x) intersection point xCIf the number of hits being calculated is greater than 1, chooses intersection point and exist
MSFk1、MSFk2Between separation of the x value as x-axis.Finally, defining distribution map regional determination parameter is DMTF=sign
(MTFk1-MTFk2), such as distribution map regional determination parameter DMTF × (x-xC) > 0, kth dimensional feature distribution map relationship isOtherwise kth dimensional feature distribution map relationship is
The present invention is further illustrated below by a specific embodiment.
Using Micro-seismic Signals as research object in the embodiment, due to Micro-seismic Signals by geological conditions influenced compared with
Greatly, the phenomenon that Micro-seismic Signals feature generated under different geological conditions differs greatly, that is, generates characteristic offset, and geology item
Part is not change over time and change.In emulation experiment, using first to signal carry out FFT transform ask again the method for cepstrum come into
Row feature extraction, intrinsic dimensionality M=511, classifier use Fuzzy Classifier then to carry out Classification and Identification, and fuzzy classification is subordinate to
It spends function and selects Gaussian function, Classification and Identification target is wheeled vehicle and two class of creeper truck, it is assumed that for training Classification and Identification coefficient
Signal characteristic from geological conditions 0, k=1 ..., 511 in each step below.
Due to fuzzy classification subordinating degree function select be Gaussian function, from trained Classification and Identification coefficient
It can be obtained by the mean value MTF for training the two every one-dimensional characteristics of class target signature of classifier coefficientk1、MTFk2With standard deviation
STDTFk1、STDTFk2。
Calculate the mean value MSF that each dimensional feature sequence of two class target of geological conditions of characteristic offset occursk1、MSFk2With mark
Quasi- difference STDSFk1、STDSFk2。
Calculate the parameter DP for determining whether overturn per one-dimensional x-axisk=(MTFk1-MTFk2)×(MSFk1-MSFk2), according to DPk
Whether less than 0, to determine, often whether one-dimensional x-axis is overturn, i.e. xk=-xk, MSFk1=-MSFk1, MSFk2=-MSFk2。
Construct the gauss of distribution function of each dimensional feature sequence of two class targets
With
Calculate two classification target distribution function DF in every one-dimensional characteristick1(xk) and DFk2(xk) intersection point xCk, it is calculated
Number of hits is likely larger than 1, chooses intersection point in MSFk1、MSFk2Between xkValue is used as xkThe separation of axis.
Calculate the parameter DMTF that distribution map region is determined in every one-dimensional characteristick=sign (MTFk1-MTFk2), per one-dimensional spy
Sign building distribution map relationship: such as DMTFk×(xk-xCk) > 0,Otherwise
Feature Mapping is carried out to the Classification and Identification feature that characteristic offset occurs according to the distribution map relationship of above-mentioned building, then
Feature after mapping is sent into classifier and carries out identification classification.
The classifying identification method of no Feature Mapping and the classifying identification method of Feature Mapping is respectively adopted to 3 kinds of generation features
The wheeled vehicle acquired under the geological conditions of offset is emulated with creeper truck vibration signal, counts 3 kinds of geological conditions of two methods
Under every classification target correct recognition rata, as shown in table 1.
1 simulation result statistical form of table
It is not difficult to find that the present invention passes through the Gaussian Profile to every one-dimensional characteristic in the case of generation characteristic offset both front and back
Situation is counted respectively, is then constructed respectively according to the distribution situation that statistics obtains to every one-dimensional characteristic in each two classifier
Distribution map relationship is corrected the feature that characteristic offset occurs by the distribution map relationship of building.This method can be
There is the case where characteristic offset effectively correction Classification and Identification feature, so as to avoid causing because of characteristic offset Classification and Identification quasi-
The case where true rate is greatly lowered.
Claims (4)
1. a kind of Classification and Identification Feature Mapping method based on two classifiers, which comprises the following steps:
(1) the Gaussian Profile situation for training the every one-dimensional characteristic of all kinds of target signatures of classifier coefficient is counted respectively；
(2) all kinds of targets in the case where counting respectively relative to all kinds of target signatures generation characteristic offset for training classifier coefficient
The Gaussian Profile situation of the every one-dimensional characteristic of feature；
(3) the Gaussian Profile situation obtained according to statistics is to every one-dimensional in two classifier each in the case of characteristic offset both front and back
Feature constructs distribution map relationship respectively.
2. the Classification and Identification Feature Mapping method according to claim 1 based on two classifiers, which is characterized in that the step
Suddenly (1) and step (2) are all made of the Gaussian Profile situation of following method statistic: defining a kind of target signature number is N, feature dimensions
Number is M, characteristic sequence F1,...,FN(Fi=[Fi1,...,FiM]), then kth dimensional feature sequence is F1k,...,FNk, calculate kth
The mean value of dimensional feature sequenceCalculate the standard deviation of kth dimensional feature sequence
3. the Classification and Identification Feature Mapping method according to claim 1 based on two classifiers, which is characterized in that the step
Suddenly (3) specifically include following sub-step:
(31) defining the mean value in two classifiers for training two class target kth dimensional feature sequences of classifier coefficient is
MTFk1And MTFk2, standard deviation STDTFk1And STDTFk2, two class target kth dimensional feature sequences in the case where characteristic offset occur
Mean value be MSFk1And MSFk2, standard deviation STDSFk1And STDSFk2；
(32) defining x-axis overturning parameter is DP=(MTFk1-MTFk2)×(MSFk1-MSFk2), if x-axis overturns parameter DP < 0, then
X-axis is overturn, i.e. x=-x, MSFk1=-MSFk1, MSFk2=-MSFk2, otherwise it is directly entered step (33)；
(33) gauss of distribution function of two class target kth dimensional feature sequences is constructed according to the mean value and standard deviation that are calculated
(34) gauss of distribution function DF is calculated1(x) and DF2(x) intersection point xC；
(35) defining distribution map regional determination parameter is DMTF=sign (MTFk1-MTFk2), as distribution map regional determination is joined
Number DMTF × (x-xC) > 0, kth dimensional feature distribution map relationship isOtherwise
K dimensional feature distribution map relationship is
4. the Classification and Identification Feature Mapping method according to claim 1 based on two classifiers, which is characterized in that the step
Suddenly intersection point is chosen in MSF if the number of hits being calculated is greater than 1 in (34)k1、MSFk2Between x value as x-axis point
Boundary's point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811512493.1A CN109740635B (en) | 2018-12-11 | 2018-12-11 | Classification recognition feature mapping method based on two classifiers |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811512493.1A CN109740635B (en) | 2018-12-11 | 2018-12-11 | Classification recognition feature mapping method based on two classifiers |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109740635A true CN109740635A (en) | 2019-05-10 |
| CN109740635B CN109740635B (en) | 2023-12-26 |
Family
ID=66358694
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811512493.1A Active CN109740635B (en) | 2018-12-11 | 2018-12-11 | Classification recognition feature mapping method based on two classifiers |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109740635B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110040711A1 (en) * | 2009-08-14 | 2011-02-17 | Xerox Corporation | Training a classifier by dimension-wise embedding of training data |
| US20120254077A1 (en) * | 2011-03-31 | 2012-10-04 | Fatih Porikli | Data Driven Frequency Mapping for Kernels Used in Support Vector Machines |
| CN103544963A (en) * | 2013-11-07 | 2014-01-29 | 东南大学 | Voice emotion recognition method based on core semi-supervised discrimination and analysis |
| CN108647602A (en) * | 2018-04-28 | 2018-10-12 | 北京航空航天大学 | A kind of aerial remote sensing images scene classification method based on image complexity judgement |
-
2018
- 2018-12-11 CN CN201811512493.1A patent/CN109740635B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110040711A1 (en) * | 2009-08-14 | 2011-02-17 | Xerox Corporation | Training a classifier by dimension-wise embedding of training data |
| US20120254077A1 (en) * | 2011-03-31 | 2012-10-04 | Fatih Porikli | Data Driven Frequency Mapping for Kernels Used in Support Vector Machines |
| CN103544963A (en) * | 2013-11-07 | 2014-01-29 | 东南大学 | Voice emotion recognition method based on core semi-supervised discrimination and analysis |
| CN108647602A (en) * | 2018-04-28 | 2018-10-12 | 北京航空航天大学 | A kind of aerial remote sensing images scene classification method based on image complexity judgement |
Non-Patent Citations (2)
| Title |
|---|
| 张亚东;华春蓉;董大伟;闫兵;鲁志文;谢逍;: "用11/2维谱结合小波包能量提取地震动信号特征", 噪声与振动控制, no. 01 * |
| 黄杰;朱广平;: "水声目标识别中的K-D树KNN-SVM分类器研究", 海洋技术学报, no. 01 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109740635B (en) | 2023-12-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Qu et al. | Radar signal intra-pulse modulation recognition based on convolutional denoising autoencoder and deep convolutional neural network | |
| CN107577999B (en) | Radar signal intra-pulse modulation mode identification method based on singular value and fractal dimension | |
| CN110197209B (en) | Radiation source identification method based on multi-feature fusion | |
| CN105044701B (en) | Ground target sorting technique based on robustness time-frequency characteristics | |
| CN105718848B (en) | Quality evaluation method and device for fingerprint image | |
| CN106483514B (en) | Airplane motion mode identification method based on EEMD and support vector machine | |
| CN107315996A (en) | A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA | |
| CN106501800A (en) | Based on tracking before the MIMO radar target detection of cost reference particle filter | |
| CN111680737B (en) | Radar radiation source individual identification method under differential signal-to-noise ratio condition | |
| CN103217676A (en) | Radar target identification method under noise background based on bispectrum de-noising | |
| CN103020645A (en) | System and method for junk picture recognition | |
| CN103994820B (en) | A kind of moving target recognition methods based on micropore diameter microphone array | |
| CN103942444A (en) | Parity 1-norm unequal length sequence similarity metric algorithm based on DTW | |
| CN113239880A (en) | Radar radiation source identification method based on improved random forest | |
| CN107479036A (en) | A kind of radar signal feature method | |
| CN109740635A (en) | Classification and Identification Feature Mapping method based on two classifiers | |
| CN110109068A (en) | Decoy jamming recognition methods based on polarization characteristic vector | |
| CN107334481B (en) | Driving distraction detection method and system | |
| CN103942526A (en) | Linear feature extraction method for discrete data point set | |
| CN108376266A (en) | One-class support vector machines Optimization Method of Kernel Parameter based on sample edge point internal point | |
| CN103810287B (en) | Based on the image classification method having the shared assembly topic model of supervision | |
| CN111832632A (en) | Radar signal sorting method and system based on high-order frequency spectrum symmetry Holder coefficient | |
| CN110988856A (en) | Target detection trace agglomeration algorithm based on density clustering | |
| CN117054998A (en) | Cluster target detection method and system based on information entropy background stability | |
| Yu et al. | Sorting radar signal based on the resemblance coefficient of bispectrum two dimensions characteristic |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |