CN109740423A - Ethnic recognition methods and system based on face and wavelet packet analysis - Google Patents
Ethnic recognition methods and system based on face and wavelet packet analysis Download PDFInfo
- Publication number
- CN109740423A CN109740423A CN201811403283.9A CN201811403283A CN109740423A CN 109740423 A CN109740423 A CN 109740423A CN 201811403283 A CN201811403283 A CN 201811403283A CN 109740423 A CN109740423 A CN 109740423A
- Authority
- CN
- China
- Prior art keywords
- face
- wavelet packet
- ethnic
- feature vector
- image
- 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
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of ethnic recognition methods and system based on face and wavelet packet analysis, method is comprising steps of selection includes the training sample set of not agnate face, to the corresponding decomposition coefficient of facial image progress WAVELET PACKET DECOMPOSITION acquisition；Its decomposition coefficient is screened, and using the proper subspace of LDB method choice face race, obtains the optimal base after wavelet package transforms；Training sample set image is projected to respectively on face racial traits subspace, the feature vector set for obtaining training sample set image is calculated separately, is trained using SVM classifier, obtains training pattern；Testing image is identified and classified using training pattern, obtaining classification results is race belonging to the test face.The present invention extracts the racial traits of face using wavelet package transforms, is classified using SVM classifier to racial traits, realizes the identification of face race, has obtained preferable ethnic recognition effect, has provided auxiliary information for personnel's identity validation.
Description
Technical field
The present invention relates to a kind of ethnic recognition methods and system based on face and wavelet packet analysis, belong to artificial intelligence skill
Art field.
Background technique
With the international development in world market, the trans-regional exchange of the people of the different colours of skin and activity are more and more common, phase
The service trade of pass also increasingly focuses on differentiation and customizes service.Service trade, such as restaurant, in face of the visitor of different races
Family needs to provide service according to different lives and eating habit, therefore identifies not agnate customer group, takes for being promoted
Business quality is played an important role with the accurately customization service that provides and meaning.In addition, accurately species not only can be with
The face characteristic in human face data is effectively obtained, more face semantic understanding information can also be obtained, help to improve people
The identity information recognition accuracy for demonstrate,proving unification, provides auxiliary information for personnel's identity validation.
Race's identification just occurs in recent years, studies usually as a sub- project of recognition of face.Recognition of face
Be generally divided into three steps: the first step is Face datection, followed by extract face characteristic, be finally by this feature compared with having deposited feature
Compared with to achieve the purpose that recognition of face.Wherein the first step is a very crucial step, and having many algorithms at present can achieve people
The purpose of face detection, (Face Recognition of Hu Zhanqi, the Liu Hongwei based on wavelet analysis and geometrical characteristic is micro- for document 1
Type machine and application, 2009；28 (15): wavelet analysis and geometrical characteristic 21-24) is combined to realize the higher face inspection of accuracy rate
Method of determining and calculating；(the Dalian research of ginger he face recognition algorithms of the based on geometrical characteristic: Dalian University of Technology master opinion of document 2
Text, 2008) disadvantage mentioned above then is overcome using classifier, but its wrong diagnosis rate is bigger, while in second step feature extraction step,
Which employs classifiers to extract 7 groups of characteristic points: the width of left eye, the vertical range of nose and eyes line, face right boundary
Distance, width, two centers and the left corners of the mouth horizontal distance of mouth, the horizontal distance in two outsides, right eye outside canthus
With the horizontal distance of nose item, the inside canthus of left eye and the horizontal distance on nose top, the vertical range at mouth midpoint and nose, nose
It is defined as at a distance from the corners of the mouth, and by these distance feature values and eyes midpoint to the ratio between the vertical range between mouth midpoint
Standardized feature vector.
After the key message of race's identification is in addition to more intuitive Skin Color Information, also with the sky of detail on face
Between feature.But document 1 requires face location to rectify, therefore is only used for detection particular picture, can not adapt to dynamic detection；
And its extracted characteristic point of document 2 is excessively careful, requires classifier high, realization difficulty.
Summary of the invention
For deficiency existing for above method, race's identification based on face and wavelet packet analysis that the invention proposes a kind of
Method and system have obtained preferable recognition effect, have been personnel identity by the way that wavelet packet analysis algorithm to be used in race's identification
Confirmation provides auxiliary information.
The present invention solves its technical problem and adopts the technical scheme that:
On the one hand, a kind of ethnic recognition methods based on face and wavelet packet analysis provided in an embodiment of the present invention, comprising:
Step 1: selection includes the training sample set of not agnate face, carries out WAVELET PACKET DECOMPOSITION to facial image and obtains phase
The decomposition coefficient answered；
Step 2: its decomposition coefficient being screened, and using the proper subspace of LDB method choice face race, is obtained
Optimal base after wavelet package transforms；
Step 3: training sample set image is projected to respectively on face racial traits subspace, calculates separately and is trained
The feature vector set of sample set image, is trained using SVM classifier, obtains training pattern；
Step 4: testing image being identified and classified using training pattern, obtaining classification results is the test face
Affiliated race.
It is combined as a kind of possible implementation of the present embodiment, the step 1 includes:
Step 11, reading in includes not agnate human face data, and the ethnic classification being currently known is black, white, yellow three kinds, if
The feature vector of three classes race's facial image is respectively ω1, ω2, ω3, training sample set are as follows:
{ωi={ ξ(i,k)},1≤i≤3,1≤k≤Mi},
For ωiIn k-th m tie up sample vector, MiFor ωiClass sample
Number；
Step 12, selection Daubechies wavelet function db4 is to ξ(i,k)Two layers of WAVELET PACKET DECOMPOSITION is carried out, is obtained in j-th stage
N-th of wavelet packet subspace is In wavelet packet coefficient be
Step 13, it calculatesIn p-th of wavelet packet coefficientCorresponding time-frequency energy Γ(y)(ωi), to wavelet packet point
The each sub-spaces solvedCorresponding each feature vector ξ(a,y), calculate its discriminate energy:
δi=ε (Γ(1)(ωi),…,Γ(y)(ωi)) (2.2)
Wherein, y ωiClassification ordinal number on direction, NyFor the quantity of such sample vector, ε is operator of minimizing.
It is combined as a kind of possible implementation of the present embodiment, the step 2 includes:
Step 21, using Mallet pyramid algorith, the discriminate energy that step 13 is obtained is as cost function, to spy
Levy the discriminate energy δ of vectoriCarry out descending arrangement；
Step 22, m discriminate energy δ is chosen from big to smalli, and select from feature vector the corresponding features of m to
Amount, this m feature vector is exactly the proper subspace of face race, the optimal base as wavelet packet.
It is combined as a kind of possible implementation of the present embodiment, the step 3 includes:
Step 31, on the proper subspace training sample set image projection to face race, training sample is calculated
Collect the feature vector set V of image, and label is done according to the feature vector that corresponding ethnic information is each sample image；
Step 32, the feature vector set V of training sample set image is input to SVM classifier to be trained, is instructed
Practice model M.
It is combined as a kind of possible implementation of the present embodiment, the step 4 includes:
Step 41, on the proper subspace image projection to be identified to face race, images to be recognized is calculated
Feature vector R；
Step 42, feature vector R is input to SVM classifier, is classified using training pattern M, to obtain wait know
Ethnic classification belonging to others' face image.
On the other hand, a kind of ethnic identifying system based on face and wavelet packet analysis provided in an embodiment of the present invention, packet
It includes:
WAVELET PACKET DECOMPOSITION module, selection include the training sample set of not agnate face, carry out wavelet packet to facial image
It decomposes and obtains corresponding decomposition coefficient；
Optimal base obtains module, screens to its decomposition coefficient, and utilizes the feature of LDB method choice face race
Space obtains the optimal base after wavelet package transforms；
Training pattern module respectively projects to training sample set image on face racial traits subspace, calculates separately
The feature vector set of training sample set image is obtained, is trained using SVM classifier, obtains training pattern；
Face recognition module is identified and is classified to testing image using training pattern, is obtained classification results and as should
Test race belonging to face.
It is combined as a kind of possible implementation of the present embodiment, the WAVELET PACKET DECOMPOSITION module includes:
Sample training module, reading in includes not agnate human face data, and the ethnic classification being currently known is black, white, yellow
Three kinds, if the feature vector of three classes race's facial image is respectively ω1, ω2, ω3, training sample set are as follows:
{ωi={ ξ(i,k)},1≤i≤3,1≤k≤Mi},
For ωiIn k-th m tie up sample vector, MiFor ωiClass sample
Number；
WAVELET PACKET DECOMPOSITION module selects Daubechies wavelet function db4 to ξ(i,k)Two layers of WAVELET PACKET DECOMPOSITION is carried out, is obtained
N-th of wavelet packet subspace is in j-th stage In wavelet packet coefficient be
Discriminate energy computation module calculatesIn p-th of wavelet packet coefficientCorresponding time-frequency energy Γ(y)
(ωi), each sub-spaces that WAVELET PACKET DECOMPOSITION is gone outCorresponding each feature vector ξ(a,y), calculate its discriminate
Energy:
δi=ε (Γ(1)(ωi),…,Γ(y)(ωi)) (2.2)
Wherein, y ωiClassification ordinal number on direction, NyFor the quantity of such sample vector, ε is operator of minimizing.
It is combined as a kind of possible implementation of the present embodiment, the optimal base obtains module and includes:
Face recognition module, using Mallet pyramid algorith, the discriminate energy that step 13 is obtained is as cost letter
Number, to the discriminate energy δ of feature vectoriCarry out descending arrangement；
Optimal base determining module chooses m discriminate energy δ from big to smalli, and select m a corresponding from feature vector
Feature vector, this m feature vector is exactly the proper subspace of face race, the optimal base as wavelet packet.
It is combined as a kind of possible implementation of the present embodiment, the training pattern module includes:
Feature vector set calculation module, on the proper subspace training sample set image projection to face race, meter
Calculation obtains the feature vector set V of training sample set image, and is the feature of each sample image according to corresponding ethnic information
Vector makees label；
The feature vector set V of training sample set image is input to SVM classifier and is trained, obtained by SVM classifier
Training pattern M.
It is combined as a kind of possible implementation of the present embodiment, the face recognition module includes:
Image projection module on the proper subspace image projection to be identified to face race, is calculated wait know
The feature vector R of other image；
Ethnic classification obtains module, and feature vector R is input to SVM classifier, is classified using training pattern M, from
And obtain ethnic classification belonging to facial image to be identified.
What the technical solution of the embodiment of the present invention can have has the beneficial effect that:
A kind of ethnic recognition methods based on face and wavelet packet analysis of the technical solution of the embodiment of the present invention, including with
Lower step: selection includes the training sample set of not agnate face, carries out WAVELET PACKET DECOMPOSITION to facial image and obtains corresponding point
Solve coefficient；Its decomposition coefficient is screened, and using the proper subspace of LDB method choice face race, obtains wavelet packet
Optimal base after transformation；Training sample set image is projected to respectively on face racial traits subspace, calculates separately to obtain
The feature vector set of training sample set image, is trained using SVM classifier, obtains training pattern；Utilize training pattern
Testing image is identified and is classified, obtaining classification results is race belonging to the test face.The present invention utilizes wavelet packet
The racial traits of face are extracted in transformation, are classified using SVM classifier to racial traits, are realized the identification of face race,
Preferable ethnic recognition effect has been obtained, has provided auxiliary information for personnel's identity validation.
Feature is decomposed and extracted to facial image the invention firstly uses 2-d discrete wavelet packet, then uses LDB
(Local DiscriminantBasis) method obtains optimal classification feature, is finally classified using SVM (SVM)
Identification.Core of the invention method is to carry out the not agnate people of two layers of wavelet package transforms acquisition using Daubechies small echo db4
The proper subspace of face can be effectively applied to the extraction and classification of face racial traits, for some various service trades or
The service that foreign guest reception department provides more in every possible way is of great significance.Such as restaurant, in face of the client of different races,
It needs to provide service according to different lives and eating habit, therefore identifies not agnate customer group, serviced for being promoted
Quality is played an important role with the accurately customization service that provides and meaning.
A kind of ethnic identifying system based on face and wavelet packet analysis of the technical solution of the embodiment of the present invention, comprising:
WAVELET PACKET DECOMPOSITION module, optimal base obtain module, training pattern module and face recognition module, include difference by selection first
The training sample set of ethnic face carries out WAVELET PACKET DECOMPOSITION to facial image and obtains corresponding decomposition coefficient；Secondly it is decomposed
Coefficient is screened, and utilizes the proper subspace of LDB method choice face race, optimal after acquisition wavelet package transforms
Base；Then training sample set image is projected to respectively on face racial traits subspace, calculates separately to obtain training sample set
The feature vector set of image, is trained using SVM classifier, obtains training pattern；Finally using training pattern to be measured
Image is identified and is classified that obtaining classification results is race belonging to the test face, is realized the identification of face race, is obtained
Preferable ethnic recognition effect has been arrived, has provided auxiliary information for personnel's identity validation.
Detailed description of the invention:
Fig. 1 is a kind of ethnic recognition methods based on face and wavelet packet analysis shown according to an exemplary embodiment
Flow chart；
Fig. 2 is three ethnic facial image sample schematic diagrames, and Fig. 2 (a) is white facial image sample schematic diagram,
Fig. 2 (b) is the facial image sample schematic diagram of black race, and Fig. 2 (c) is the facial image sample schematic diagram of yellow；
Fig. 3 be a kind of image wavelet shown according to an exemplary embodiment double-bag decomposition tree hair schematic diagram；
Fig. 4 is a kind of ethnic identifying system based on face and wavelet packet analysis shown according to an exemplary embodiment
Schematic diagram.
Specific embodiment
The present invention will be further described with embodiment with reference to the accompanying drawing:
In order to clarify the technical characteristics of the invention, below by specific embodiment, and its attached drawing is combined, to this hair
It is bright to be described in detail.Following disclosure provides many different embodiments or example is used to realize different knots of the invention
Structure.In order to simplify disclosure of the invention, hereinafter the component of specific examples and setting are described.In addition, the present invention can be with
Repeat reference numerals and/or letter in different examples.This repetition is that for purposes of simplicity and clarity, itself is not indicated
Relationship between various embodiments and/or setting is discussed.It should be noted that illustrated component is not necessarily to scale in the accompanying drawings
It draws.Present invention omits the descriptions to known assemblies and treatment technology and process to avoid the present invention is unnecessarily limiting.
The extraction of facial detail feature generallys use wavelet transformation, in the analysis and different scale of Lai Shixian multiresolution
Feature extraction.Wavelet packet analysis is grown up on the basis of wavelet analysis, is the popularization of wavelet analysis, it is capable of providing
A kind of finer analysis method, it can further decompose the high frequency section that wavelet analysis does not segment, and energy
Enough according to the corresponding frequency band of the adaptive selection of the feature of analyzed signal, it is allowed to match with signal spectrum, to improve
Time frequency resolution has the advantages that the main information that can retain image retains different directions detailed information again, therefore, in race
When feature extraction, wavelet packet analysis has more application value and practicability, and realizes the key technology of race's identification.
This patent extracts wavelet packet most by carrying out db4 wavelet package transforms to not agnate face sample set image
Training sample image collection is mapped to this feature subspace, passed through by excellent base using the proper subspace of LDB selection face race
Calculate the characteristic vector set that can be obtained by training sample set image.According to the ethnic classification of sample, to characteristic vector set
Label is made, is trained using SVM classifier, the disaggregated model of face race is obtained.It can be carried out below using the model
The species of face identify.It is people by wavelet packet analysis algorithm for having obtained preferable recognition effect in race's identification
Member's identity validation provides auxiliary information.
Fig. 1 is a kind of ethnic recognition methods based on face and wavelet packet analysis shown according to an exemplary embodiment
Flow chart.As shown in Figure 1, a kind of ethnic recognition methods based on face and wavelet packet analysis provided in an embodiment of the present invention, packet
It includes:
Step 1: selection includes the training sample set of not agnate face, carries out WAVELET PACKET DECOMPOSITION to facial image, obtains
Corresponding decomposition coefficient.
The step 1 specifically includes:
Step 11, reading in includes not agnate human face data, and the ethnic classification being currently known is black, white, yellow three kinds, such as
Shown in Fig. 2, if the feature vector of three classes race's facial image is respectively ω1, ω2, ω3, training sample set are as follows:
{ωi={ ξ(i,k)},1≤i≤3,1≤k≤Mi},
For ωiIn k-th m tie up sample vector, MiFor ωiClass sample
Number；
Step 12, selection Daubechies wavelet function db4 is to ξ(i,k)Two layers of WAVELET PACKET DECOMPOSITION is carried out, is obtained in j-th stage
N-th of wavelet packet subspace is In wavelet packet coefficient be
Step 13, it calculatesIn p-th of wavelet packet coefficientCorresponding time-frequency energy Γ(y)(ωi), to wavelet packet point
The each sub-spaces solvedCorresponding each feature vector ξ(a,y), calculate its discriminate energy:
δi=ε (Γ(1)(ωi),…,Γ(y)(ωi)) (2.2)
Wherein, y ωiClassification ordinal number on direction, NyFor the quantity of such sample vector, ε is operator of minimizing.
This patent passes through a given function u using Daubechies small echo db4 as basic wavelet mother function Ψ0
(t), wavelet packet basis functions race is generated using following formula
Wherein, hzAnd gkIt is the filter coefficient of Orthogonal Wavelets, function u0It (t) can be female with scaling function φ and small echo
Function Ψ definition,It is all
Function set { the u constructed according to above-mentioned formulan(t)}n∈ZIt is exactly by u0Wavelet packet determined by=φ, that is, this patent
For carrying out the basic function set of feature extraction.
The wavelet packet basis functions set constructed using Daubechies small echo db4, so that it may for facial image according to Fig. 3
It is shown to be decomposed.In Fig. 3, A indicates that low frequency, D indicate high frequency, and the serial number at end indicates the number of plies of WAVELET PACKET DECOMPOSITION, i.e. ruler
Degree.
Step 2: its decomposition coefficient being screened, and using the proper subspace of LDB method choice face race, is obtained
Optimal base after wavelet package transforms.
The step 2 specifically includes:
Step 21, using Mallet pyramid algorith, the discriminate energy that step 13 is obtained is as cost function, to spy
Levy the discriminate energy δ of vectoriCarry out descending arrangement；
Step 22, m discriminate energy δ is chosen from big to smalli, and select from feature vector the corresponding features of m to
Amount, this m feature vector is exactly the proper subspace of face race, the optimal base as wavelet packet.
Step 3: training sample set image is projected to respectively on face racial traits subspace, calculates separately and is trained
The feature vector set of sample set image, is trained using SVM classifier, obtains training pattern.
The step 3 specifically includes:
Step 31, on the proper subspace training sample set image projection to face race, training sample is calculated
Collect the feature vector set V of image, and label is done according to the feature vector that corresponding ethnic information is each sample image；
Step 32, the feature vector set V of training sample set image is input to SVM classifier to be trained, is instructed
Practice model M.
Step 4: testing image being identified and classified using training pattern, obtaining classification results is the test face
Affiliated race.
The step 4 specifically includes:
Step 41, on the proper subspace image projection to be identified to face race, images to be recognized is calculated
Feature vector R；
Step 42, feature vector R is input to SVM classifier, is classified using training pattern M, to obtain wait know
Ethnic classification belonging to others' face image.
The present embodiment extracts the racial traits of face using wavelet package transforms, is carried out using SVM classifier to racial traits
Classification, realizes the identification of face race, has obtained preferable ethnic recognition effect, provides auxiliary letter for personnel's identity validation
Breath.
Fig. 4 is a kind of ethnic identifying system based on face and wavelet packet analysis shown according to an exemplary embodiment
Schematic diagram.As shown in figure 4, a kind of ethnic identifying system based on face and wavelet packet analysis provided in this embodiment includes:
WAVELET PACKET DECOMPOSITION module, selection include the training sample set of not agnate face, carry out wavelet packet to facial image
It decomposes and obtains corresponding decomposition coefficient；
Optimal base obtains module, screens to its decomposition coefficient, and utilizes the feature of LDB method choice face race
Space obtains the optimal base after wavelet package transforms；
Training pattern module respectively projects to training sample set image on face racial traits subspace, calculates separately
The feature vector set of training sample set image is obtained, is trained using SVM classifier, obtains training pattern；
Face recognition module is identified and is classified to testing image using training pattern, is obtained classification results and as should
Test race belonging to face.
As a kind of possible implementation, the WAVELET PACKET DECOMPOSITION module includes:
Sample training module, reading in includes not agnate human face data, and the ethnic classification being currently known is black, white, yellow
Three kinds, if the feature vector of three classes race's facial image is respectively ω1, ω2, ω3, training sample set are as follows:
{ωi={ ξ(i,k)},1≤i≤3,1≤k≤Mi},
For ωiIn k-th m tie up sample vector, MiFor ωiClass sample
Number；
WAVELET PACKET DECOMPOSITION module selects Daubechies wavelet function db4 to ξ(i,k)Two layers of WAVELET PACKET DECOMPOSITION is carried out, is obtained
N-th of wavelet packet subspace is in j-th stage In wavelet packet coefficient be
Discriminate energy computation module calculatesIn p-th of wavelet packet coefficientCorresponding time-frequency energy Γ(y)
(ωi), each sub-spaces that WAVELET PACKET DECOMPOSITION is gone outCorresponding each feature vector ξ(a,y), calculate its discriminate
Energy:
δi=ε (Γ(1)(ωi),…,Γ(y)(ωi)) (2.2)
Wherein, y ωiClassification ordinal number on direction, NyFor the quantity of such sample vector, ε is operator of minimizing.
As a kind of possible implementation, the optimal base obtains module and includes:
Face recognition module, using Mallet pyramid algorith, the discriminate energy that step 13 is obtained is as cost letter
Number, to the discriminate energy δ of feature vectoriCarry out descending arrangement；
Optimal base determining module chooses m discriminate energy δ from big to smalli, and select m a corresponding from feature vector
Feature vector, this m feature vector is exactly the proper subspace of face race, the optimal base as wavelet packet.
As a kind of possible implementation, the training pattern module includes:
Feature vector set calculation module, on the proper subspace training sample set image projection to face race, meter
Calculation obtains the feature vector set V of training sample set image, and is the feature of each sample image according to corresponding ethnic information
Vector makees label；
The feature vector set V of training sample set image is input to SVM classifier and is trained, obtained by SVM classifier
Training pattern M.
As a kind of possible implementation, the face recognition module includes:
Image projection module on the proper subspace image projection to be identified to face race, is calculated wait know
The feature vector R of other image；
Ethnic classification obtains module, and feature vector R is input to SVM classifier, is classified using training pattern M, from
And obtain ethnic classification belonging to facial image to be identified.
The present embodiment passes through the training sample set that selection includes not agnate face first, carries out wavelet packet to facial image
It decomposes and obtains corresponding decomposition coefficient；Secondly its decomposition coefficient is screened, and utilizes the spy of LDB method choice face race
Subspace is levied, the optimal base after wavelet package transforms is obtained；Then it is special training sample set image to be projected to face race respectively
It levies on subspace, calculates separately the feature vector set for obtaining training sample set image, be trained, obtained using SVM classifier
To training pattern；Finally testing image is identified and classified using training pattern, obtaining classification results is the tester
Race belonging to face realizes the identification of face race, has obtained preferable ethnic recognition effect, has provided for personnel's identity validation
Auxiliary information.
A kind of ethnic recognition methods based on face and wavelet packet analysis of the invention is first with 2-d discrete wavelet packet
Feature is decomposed and extracted to facial image, then is obtained most preferably using LDB (Local DiscriminantBasis) method
Characteristic of division finally carries out Classification and Identification using SVM (SVM).
Core of the invention method is not agnate using Daubechies small echo db4 two layers of wavelet package transforms acquisition of progress
The proper subspace of face can be effectively applied to the extraction and classification of face racial traits, for some various service trades or
The service that person foreign guest reception department provides more in every possible way is of great significance.Such as restaurant, in face of the visitor of different races
Family needs to provide service according to different lives and eating habit, therefore identifies not agnate customer group, takes for being promoted
Business quality is played an important role with the accurately customization service that provides and meaning.
The above is the preferred embodiment of the present invention, for those skilled in the art,
Without departing from the principles of the invention, several improvements and modifications can also be made, these improvements and modifications are also regarded as this
The protection scope of invention.
Claims (10)
1. a kind of ethnic recognition methods based on face and wavelet packet analysis characterized by comprising
Step 1: selection includes the training sample set of not agnate face, carries out WAVELET PACKET DECOMPOSITION to facial image and obtains accordingly
Decomposition coefficient；
Step 2: its decomposition coefficient being screened, and using the proper subspace of LDB method choice face race, obtains small echo
Optimal base after packet transform；
Step 3: training sample set image is projected to respectively on face racial traits subspace, calculates separately to obtain training sample
The feature vector set for collecting image, is trained using SVM classifier, obtains training pattern；
Step 4: testing image being identified and classified using training pattern, obtaining classification results is belonging to the test face
Race.
2. the ethnic recognition methods according to claim 1 based on face and wavelet packet analysis, which is characterized in that the step
Rapid 1 includes:
Step 11, reading in includes not agnate human face data, and the ethnic classification being currently known is black, white, yellow three kinds, if three classes
The feature vector of ethnic facial image is respectively ω1, ω2, ω3, training sample set are as follows:
{ωi={ ξ(i,k)},1≤i≤3,1≤k≤Mi},
For ωiIn k-th m tie up sample vector, MiFor ωiClass sample number；
Step 12, selection Daubechies wavelet function db4 is to ξ(i,k)Two layers of WAVELET PACKET DECOMPOSITION is carried out, is obtained in j-th stage n-th
Wavelet packet subspace is In wavelet packet coefficient be
Step 13, it calculatesIn p-th of wavelet packet coefficientCorresponding time-frequency energy Γ(y)(ωi), WAVELET PACKET DECOMPOSITION is gone out
Each sub-spacesCorresponding each feature vector ξ(a,y), calculate its discriminate energy:
δi=ε (Γ(1)(ωi),…,Γ(y)(ωi)) (2.2)
Wherein, y ωiClassification ordinal number on direction, NyFor the quantity of such sample vector, ε is operator of minimizing.
3. the ethnic recognition methods according to claim 2 based on face and wavelet packet analysis, it is characterised in that: the step
Rapid 2 include:
Step 21, using Mallet pyramid algorith, the discriminate energy that step 13 is obtained as cost function, to feature to
The discriminate energy δ of amountiCarry out descending arrangement；
Step 22, m discriminate energy δ is chosen from big to smalli, and m corresponding feature vectors, this m are selected from feature vector
A feature vector is exactly the proper subspace of face race, the optimal base as wavelet packet.
4. the ethnic recognition methods according to claim 1 based on face and wavelet packet analysis, which is characterized in that the step
Rapid 3 include:
Step 31, on the proper subspace training sample set image projection to face race, training sample set figure is calculated
The feature vector set V of picture, and label is done according to the feature vector that corresponding ethnic information is each sample image；
Step 32, the feature vector set V of training sample set image is input to SVM classifier to be trained, obtains training mould
Type M.
5. the ethnic recognition methods according to any one of claims 1-4 based on face and wavelet packet analysis, feature
Be: the step 4 includes:
Step 41, on the proper subspace image projection to be identified to face race, the spy of images to be recognized is calculated
Levy vector R；
Step 42, feature vector R is input to SVM classifier, is classified using training pattern M, to obtain people to be identified
Ethnic classification belonging to face image.
6. a kind of ethnic identifying system based on face and wavelet packet analysis characterized by comprising
WAVELET PACKET DECOMPOSITION module, selection include the training sample set of not agnate face, carry out WAVELET PACKET DECOMPOSITION to facial image
Obtain corresponding decomposition coefficient；
Optimal base obtains module, screens to its decomposition coefficient, and empty using the feature of LDB method choice face race
Between, obtain the optimal base after wavelet package transforms；
Training pattern module respectively projects to training sample set image on face racial traits subspace, calculates separately to obtain
The feature vector set of training sample set image, is trained using SVM classifier, obtains training pattern；
Face recognition module identified and classified to testing image using training pattern, and obtaining classification results is the test
Race belonging to face.
7. the ethnic identifying system according to claim 6 based on face and wavelet packet analysis, which is characterized in that described small
Wave packet decomposing module includes:
Sample training module, reading in includes not agnate human face data, and the ethnic classification being currently known is black, white, yellow three kinds,
If the feature vector of three classes race's facial image is respectively ω1, ω2, ω3, training sample set are as follows:
{ωi={ ξ(i,k)},1≤i≤3,1≤k≤Mi},
For ωiIn k-th m tie up sample vector, MiFor ωiClass sample number；
WAVELET PACKET DECOMPOSITION module selects Daubechies wavelet function db4 to ξ(i,k)Two layers of WAVELET PACKET DECOMPOSITION is carried out, jth is obtained
N-th of wavelet packet subspace is in grade In wavelet packet coefficient be
Discriminate energy computation module calculatesIn p-th of wavelet packet coefficientCorresponding time-frequency energy Γ(y)(ωi), it is right
Each sub-spaces that WAVELET PACKET DECOMPOSITION goes outCorresponding each feature vector ξ(a,y), calculate its discriminate energy:
δi=ε (Γ(1)(ωi),…,Γ(y)(ωi)) (2.2)
Wherein, y ωiClassification ordinal number on direction, NyFor the quantity of such sample vector, ε is operator of minimizing.
8. the ethnic identifying system according to claim 7 based on face and wavelet packet analysis, which is characterized in that it is described most
Excellent base obtains module
Face recognition module, using Mallet pyramid algorith, the discriminate energy that step 13 is obtained is right as cost function
The discriminate energy δ of feature vectoriCarry out descending arrangement；
Optimal base determining module chooses m discriminate energy δ from big to smalli, and m corresponding features are selected from feature vector
Vector, this m feature vector are exactly the proper subspace of face race, the optimal base as wavelet packet.
9. the ethnic identifying system according to claim 6 based on face and wavelet packet analysis, which is characterized in that the instruction
Practicing model module includes:
Feature vector set calculation module on the proper subspace training sample set image projection to face race, calculates
It is the feature vector of each sample image to the feature vector set V of training sample set image, and according to corresponding ethnic information
Make label；
The feature vector set V of training sample set image is input to SVM classifier and is trained, trained by SVM classifier
Model M.
10. the ethnic identifying system according to claim 6-9 any one based on face and wavelet packet analysis, feature
It is, the face recognition module includes:
On the proper subspace image projection to be identified to face race, figure to be identified is calculated in image projection module
The feature vector R of picture；
Ethnic classification obtains module, and feature vector R is input to SVM classifier, is classified using training pattern M, thus
To ethnic classification belonging to facial image to be identified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811403283.9A CN109740423B (en) | 2018-11-22 | 2018-11-22 | Ethnicity identification method and system based on face and wavelet packet analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811403283.9A CN109740423B (en) | 2018-11-22 | 2018-11-22 | Ethnicity identification method and system based on face and wavelet packet analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109740423A true CN109740423A (en) | 2019-05-10 |
| CN109740423B CN109740423B (en) | 2023-04-07 |
Family
ID=66358143
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811403283.9A Active CN109740423B (en) | 2018-11-22 | 2018-11-22 | Ethnicity identification method and system based on face and wavelet packet analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109740423B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110956079A (en) * | 2019-10-12 | 2020-04-03 | 深圳壹账通智能科技有限公司 | Face recognition model construction method and device, computer equipment and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101271521A (en) * | 2008-05-13 | 2008-09-24 | 清华大学 | Human face recognition method based on anisotropic double-tree complex wavelet package transforms |
| US20080253625A1 (en) * | 2006-10-10 | 2008-10-16 | Stephanie Schuckers | Multi-Resolutional Texture Analysis Fingerprint Liveness Systems and Methods |
| CN103093234A (en) * | 2012-12-21 | 2013-05-08 | 中国科学院合肥物质科学研究院 | Identity recognition method based on ground reactive force during walking |
| CN103164689A (en) * | 2011-12-16 | 2013-06-19 | 上海移远通信技术有限公司 | Face recognition method and face recognition system |
| CN104036247A (en) * | 2014-06-11 | 2014-09-10 | 杭州巨峰科技有限公司 | Facial feature based face racial classification method |
| CN105160359A (en) * | 2015-09-17 | 2015-12-16 | 华南理工大学 | Complex structure damage cooperative identification method based on ultrasonic guided-wave |
| WO2016176887A1 (en) * | 2015-05-06 | 2016-11-10 | 福州大学 | Animal sound identification method based on double spectrogram features |
| CN106650606A (en) * | 2016-10-21 | 2017-05-10 | 江苏理工学院 | Matching and processing method of face image and face image model construction system |
| CN106874920A (en) * | 2015-12-10 | 2017-06-20 | 北京航天长峰科技工业集团有限公司 | License plate character recognition method based on wavelet packet analysis and SVMs |
| CN108629376A (en) * | 2018-05-10 | 2018-10-09 | 吉林省日月智感互联科技有限公司 | A kind of microseism sensing region invasive biology method and system |
-
2018
- 2018-11-22 CN CN201811403283.9A patent/CN109740423B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080253625A1 (en) * | 2006-10-10 | 2008-10-16 | Stephanie Schuckers | Multi-Resolutional Texture Analysis Fingerprint Liveness Systems and Methods |
| CN101271521A (en) * | 2008-05-13 | 2008-09-24 | 清华大学 | Human face recognition method based on anisotropic double-tree complex wavelet package transforms |
| CN103164689A (en) * | 2011-12-16 | 2013-06-19 | 上海移远通信技术有限公司 | Face recognition method and face recognition system |
| CN103093234A (en) * | 2012-12-21 | 2013-05-08 | 中国科学院合肥物质科学研究院 | Identity recognition method based on ground reactive force during walking |
| CN104036247A (en) * | 2014-06-11 | 2014-09-10 | 杭州巨峰科技有限公司 | Facial feature based face racial classification method |
| WO2016176887A1 (en) * | 2015-05-06 | 2016-11-10 | 福州大学 | Animal sound identification method based on double spectrogram features |
| CN105160359A (en) * | 2015-09-17 | 2015-12-16 | 华南理工大学 | Complex structure damage cooperative identification method based on ultrasonic guided-wave |
| CN106874920A (en) * | 2015-12-10 | 2017-06-20 | 北京航天长峰科技工业集团有限公司 | License plate character recognition method based on wavelet packet analysis and SVMs |
| CN106650606A (en) * | 2016-10-21 | 2017-05-10 | 江苏理工学院 | Matching and processing method of face image and face image model construction system |
| CN108629376A (en) * | 2018-05-10 | 2018-10-09 | 吉林省日月智感互联科技有限公司 | A kind of microseism sensing region invasive biology method and system |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110956079A (en) * | 2019-10-12 | 2020-04-03 | 深圳壹账通智能科技有限公司 | Face recognition model construction method and device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109740423B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104866829B (en) | A kind of across age face verification method based on feature learning | |
| CN104143079B (en) | The method and system of face character identification | |
| CN109670429A (en) | A kind of the monitor video multiple target method for detecting human face and system of Case-based Reasoning segmentation | |
| CN100440246C (en) | Positioning method for human face characteristic point | |
| CN109886273A (en) | A kind of CMR classification of image segmentation system | |
| CN103324944B (en) | A kind of based on SVM with the false fingerprint detection method of rarefaction representation | |
| CN101739555B (en) | Method and system for detecting false face, and method and system for training false face model | |
| CN106980852B (en) | Based on Corner Detection and the medicine identifying system matched and its recognition methods | |
| CN110018524A (en) | A kind of X-ray safety check contraband recognition methods of view-based access control model-attribute | |
| CN106023220A (en) | Vehicle exterior part image segmentation method based on deep learning | |
| CN107924472A (en) | Pass through the image classification of brain computer interface | |
| CN105956570B (en) | Smiling face's recognition methods based on lip feature and deep learning | |
| CN103034838A (en) | Special vehicle instrument type identification and calibration method based on image characteristics | |
| CN103077378B (en) | Contactless face recognition algorithms based on extension eight neighborhood Local textural feature and system of registering | |
| CN109214336A (en) | A kind of vehicle window marker detection method and device | |
| CN103971106A (en) | Multi-view human facial image gender identification method and device | |
| CN109815979A (en) | A kind of weak label semantic segmentation nominal data generation method and system | |
| CN106203256A (en) | A kind of low resolution face identification method based on sparse holding canonical correlation analysis | |
| CN106897669A (en) | A kind of pedestrian based on consistent iteration various visual angles transfer learning discrimination method again | |
| CN106650606A (en) | Matching and processing method of face image and face image model construction system | |
| CN106529377A (en) | Age estimating method, age estimating device and age estimating system based on image | |
| CN103065126B (en) | Re-identification method of different scenes on human body images | |
| CN109508692A (en) | A kind of gender identification method and system based on 3D fingerprint image | |
| CN109876416A (en) | A kind of rope skipping method of counting based on image information | |
| CN110348320A (en) | A kind of face method for anti-counterfeit based on the fusion of more Damage degrees |
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 |