CN102194107A - Smiling face recognition method for reducing dimension by using improved linear discriminant analysis - Google Patents

Smiling face recognition method for reducing dimension by using improved linear discriminant analysis Download PDF

Info

Publication number
CN102194107A
CN102194107A CN2011101229368A CN201110122936A CN102194107A CN 102194107 A CN102194107 A CN 102194107A CN 2011101229368 A CN2011101229368 A CN 2011101229368A CN 201110122936 A CN201110122936 A CN 201110122936A CN 102194107 A CN102194107 A CN 102194107A
Authority
CN
China
Prior art keywords
sample
class
gabor
discriminant analysis
linear discriminant
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
Application number
CN2011101229368A
Other languages
Chinese (zh)
Other versions
CN102194107B (en
Inventor
郭礼华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN 201110122936 priority Critical patent/CN102194107B/en
Publication of CN102194107A publication Critical patent/CN102194107A/en
Application granted granted Critical
Publication of CN102194107B publication Critical patent/CN102194107B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention provides a smiling face face recognition method for reducing dimension by using improved linear discriminant analysis, and the method comprises the following specific steps: positioning a human face region based on a standard Haar feature and AdaBoost algorithm; extracting a human face region Gabor feature vector; performing the feature dimension reducing on the Gabor feature vector by using the improved linear discriminant analysis; and supporting the training and recognition of a vector machine. The improved linear discriminant analysis can be used for performing the feature dimension reducing, the multi-modal problem in the traditional linear discriminant analysis is overcome, and the dimension reducing of the feature dimension number can be effectively realized, in general, the smiling face recognition method has the advantages of rapid system recognition speed and high system recognition performance.

Description

A kind of smiling face's recognition methods that improves the linear discriminant analysis dimensionality reduction
Technical field
The present invention relates to pattern-recognition and field of artificial intelligence, particularly a kind of smiling face's recognition methods that improves the linear discriminant analysis dimensionality reduction.
Background technology
Smiling face's recognition technology is meant and identifies the static facial image of a width of cloth, or whether the people's face in one section dynamic video is a kind of technology that the smiling face expresses one's feelings.Smiling face's recognition technology belongs to pattern-recognition and field of artificial intelligence, it can be interpreted as a subclass of human face expression technology.Human face expression identification is meant that generally the expression with the mankind is divided into seven classes (neutrality, happy, fear that sadness, anger is detested, and disdains), classifies automatically to the human face expression in static images or the video.But the Expression Recognition technology is ripe not enough, also fails really to use in practice.Comparatively speaking, the easier realization of smiling face's recognition technology also has more practicality (except neutrality, the smiling face is the human expression that is apt to do most), so has occurred the application of some smiling face's recognition technologies on the market.For example, be the city of asking of the digital camera of attraction and camcorder with smiling face's recognition function, the typical case who is exactly smiling face's recognition technology uses.
The smiling face is one of a human face expression crucial expression, and its identification also is one importantly has a problem to be solved.Application number provides a kind of smiling face's method for tracing for the patent application document of CN200710203477.X at present, and this method is by the lens shooting picture of camera model; Convert picture shot to black and white picture; Read the brightness value of each point on the described black and white picture; By the head portrait figure in the elementary contour intercepting picture of brightness value and face, and in the head portrait figure of above-mentioned intercepting, intercept the lip figure; The similarity of lip figure realized smiling face's difference by this similarity when people's face that the lip figure of more above-mentioned intercepting and camera model are preserved was smiled, thereby determined whether picture shot exists the camera the inside.The method is only used the method for template matches, have only when the personage that takes the photograph of institute in database, the method is just very effective, and the method depends on threshold value, needs good experience judgement.Application number is the method that the patent application document of CN200710173678.x has been introduced a kind of capturing smiling face by mobile phone camera, and this method overcomes uses mobile phone photograph to be difficult for catching the defective of best right moment for camera and the best expression of one be shooted in the prior art.At first detect and obtain people's face, extract the position parameter data of people's face vitals by people's face; People's face implement is followed the tracks of, is analyzed the center of mouth, when the mouth that detects the people when laughing at, Focus Club automatically locks face, takes.This invention just utilizes the smiling face to carry out auxiliary camera and takes, and smiling face detection method is too simple, only satisfies some simple application scenarios.
Summary of the invention
The objective of the invention is to overcome the deficiency of above-mentioned smiling face's recognition technology, a kind of smiling face's recognition methods that improves the linear discriminant analysis dimensionality reduction is provided.
The technical solution used in the present invention is: use the Gabor feature in smiling face's identification and utilize the improvement linear discriminant analysis to carry out the feature dimensionality reduction; Finally, the support vector machine sorter is realized smiling face and non-smiling face's difference,
Purpose of the present invention is achieved through the following technical solutions:
(11) locate based on the human face region of Haar feature and AdaBoost algorithm;
(12) extraction of human face region Gabor eigenvector;
(13) improve the feature dimensionality reduction of linear discriminant analysis to human face region Gabor eigenvector;
(14) the Gabor eigenvector of support vector machine after to dimensionality reduction trained and discerned.
Described improvement linear discriminant analysis is to the feature dimensionality reduction of human face region Gabor eigenvector, and its concrete steps are as follows: suppose that the subspace is by a series of vectors (
Figure 113284DEST_PATH_IMAGE002
) form, these vectors are formed matrix
Figure 276280DEST_PATH_IMAGE003
, the Gabor eigenvector is a sample in each human face region, in order to improve the method for conventional linear discriminatory analysis, the present invention defines the between class scatter matrix
Figure 816983DEST_PATH_IMAGE004
With divergence matrix in the class
Figure 460454DEST_PATH_IMAGE005
For:
Figure 733304DEST_PATH_IMAGE006
Figure 462225DEST_PATH_IMAGE007
Figure 727990DEST_PATH_IMAGE008
Be the classification sum of sample,
Figure 732036DEST_PATH_IMAGE010
Expression training sample quantity,
Figure 885936DEST_PATH_IMAGE011
Be
Figure 768442DEST_PATH_IMAGE012
The number of sample in the class,
Figure 307876DEST_PATH_IMAGE013
Be the weighted value of sample,
Figure 250425DEST_PATH_IMAGE014
Be
Figure 688359DEST_PATH_IMAGE015
The average of all training samples of class, The average of representing all training samples, Be
Figure 554050DEST_PATH_IMAGE012
In the class Individual training sample,
Figure 703588DEST_PATH_IMAGE019
Be illustrated in the non-j sample in the i class; Utilize formula again
Figure 968348DEST_PATH_IMAGE020
Ask for the projection matrix of optimum improvement linear discriminant analysis
Figure 252699DEST_PATH_IMAGE021
Utilize projection matrix Training sample is carried out projection, and obtaining eigenvector is exactly primitive character.
Described Gabor feature is a kind of textural characteristics, and the Gabor feature is a kind of feature extraction algorithm of very standard, when extracting the Gabor feature, the Gabor wave filter have 4 directions (
Figure 614596DEST_PATH_IMAGE022
) and 5 yardsticks (
Figure 796178DEST_PATH_IMAGE023
).
Described support vector machine adopts two class support vector machine sorters proper vector is trained and to discern.
Human face region location and feature extraction institute at be test pattern, and training and identification division are at the concentrated all images of training sample.Training no longer needs to carry out the calculating of projection matrix with the improvement linear discriminant analysis of identification division, but the projection matrix that obtains when directly utilizing feature extraction, the Gabor feature is multiplied by projection matrix, thereby with the feature of Feature Conversion to the discriminating subspace, differentiating intercepting low-dimensional part on the subspace, reach the feature selecting purpose.
Compare with existing smiling face's recognition methods, the present invention has the following advantages:
(1) improvement linear discriminant analysis of the present invention carries out effective dimensionality reduction to feature, and the feature of low-dimensional is in sorter training and testing process, and easily realization and training time reduce greatly;
(2) improvement linear discriminant analysis of the present invention has redefined the between class scatter matrix S bWith divergence matrix S in the class w, overcome the multi-modal shortcoming of conventional linear discriminatory analysis;
(3) the present invention has selected most popular two class support vector machine sorters for use, is carrying out effective compromise on the statistical error He on the popularization ability, so under the situation of a small amount of training sample, sorter still can keep good recognition performance.
Description of drawings
Fig. 1 is a system construction drawing of the present invention.
Embodiment
The present invention is described further below in conjunction with accompanying drawing, implement the used identification equipment of the present invention and can adopt the camera collection picture, discern with computing machine, show final display result with pure flat escope, can adopt the C language to work out all kinds of handling procedures, just can implement the present invention preferably.
System architecture diagram of the present invention as shown in Figure 1, system is divided into training and identification two parts, the training part at first utilizes 10,000 front face images of camera collection to form training sample set, comprising 5000 smiling faces and 5000 non-smiling face's images.
Systematic training mainly contains following step:
1) adopt the level and smooth and histogram equalizing method of Gauss to carry out the image pre-service, removal causes noise and luminance difference in the image because of camera;
2) adopt the class Haar feature of standard and people's face detection algorithm of AdaBoost to seek human face region;
3) extraction of the Gabor eigenvector of human face region, Gabor wave filter have 4 directions (
Figure 189113DEST_PATH_IMAGE022
) and 5 yardsticks (
Figure 88936DEST_PATH_IMAGE023
);
4) with improving the linear discriminant analysis method, suppose that the subspace is by a series of vectors to Gabor feature dimensionality reduction (
Figure 878087DEST_PATH_IMAGE024
) open into, these vectors are formed matrix
Figure 441923DEST_PATH_IMAGE025
, the Gabor eigenvector is a sample in each human face region, definition between class scatter matrix
Figure 829042DEST_PATH_IMAGE026
With divergence matrix in the class
Figure 933264DEST_PATH_IMAGE027
For:
Figure 276390DEST_PATH_IMAGE028
Figure 73445DEST_PATH_IMAGE029
Figure 619964DEST_PATH_IMAGE030
Wherein Be divergence matrix in the class,
Figure 804137DEST_PATH_IMAGE032
Be the between class scatter matrix,
Figure 959044DEST_PATH_IMAGE009
Be the classification sum (C=2) of sample,
Figure 55176DEST_PATH_IMAGE010
Expression training sample quantity (n=10000),
Figure 438884DEST_PATH_IMAGE011
Be
Figure 569651DEST_PATH_IMAGE012
The number of sample in the class (
Figure 895459DEST_PATH_IMAGE033
=5000,
Figure 478887DEST_PATH_IMAGE034
=5000),
Figure 728603DEST_PATH_IMAGE013
Be the weighted value of sample, calculate acquisition by formula,
Figure 651560DEST_PATH_IMAGE035
Be all
Figure 961318DEST_PATH_IMAGE012
The average of class training sample,
Figure 218993DEST_PATH_IMAGE036
The average of representing all training samples, Be In the class
Figure 530523DEST_PATH_IMAGE018
Individual training sample,
Figure 822964DEST_PATH_IMAGE019
Be illustrated in other sample of non-j sample in the i class.Finally utilize formula
Figure 867012DEST_PATH_IMAGE037
Ask for the projection matrix of optimum improvement linear discriminant analysis
Figure 561299DEST_PATH_IMAGE021
Utilize projection matrix
Figure 212860DEST_PATH_IMAGE021
Training sample is carried out projection, and obtaining eigenvector is exactly the discriminating subspace of primitive character, differentiating intercepting low-dimensional part on the subspace, finishes the feature dimensionality reduction;
5) will improve characteristic behind the linear discriminant analysis dimensionality reduction sends into the support vector machine sorter and trains.
Enter system identification, utilize camera to gather the direct picture of people's face in real time, also through image pre-service, human face region detection, people's face Gabor feature extraction, improve the feature selecting of linear discriminant analysis method and the identification of support vector machine sorter then.Directly utilize training department to divide the projection matrix that obtains, feature is multiplied by projection matrix, realize,, reach the feature selecting purpose differentiating intercepting low-dimensional part on the subspace with the feature of Feature Conversion to the discriminating subspace; Utilize the interphase of the support vector machine sorter that trains then, judge the classification of current test data, judge that promptly current test pattern is smiling face or non-smiling face.

Claims (4)

1. smiling face's recognition methods that improves the linear discriminant analysis dimensionality reduction, it is characterized in that: concrete steps are:
(11), the human face region of measured Haar feature and AdaBoost algorithm location;
(12), the extraction of human face region Gabor eigenvector;
(13), improve the feature dimensionality reduction of linear discriminant analysis to human face region Gabor eigenvector;
(14), the Gabor eigenvector of support vector machine after to dimensionality reduction trained and discerned.
2. 1 described smiling face's recognition methods as requested is characterized in that: the Gabor eigenvector of described step (12) extracts the Gabor wave filter that adopts and has 4 directions:
Figure 253489DEST_PATH_IMAGE001
With 5 yardsticks:
3. 1 described smiling face's recognition methods as requested, it is characterized in that: the concrete steps of described step (13) are: the definition subspace is by a series of vectors
Figure 998908DEST_PATH_IMAGE003
(
Figure 51047DEST_PATH_IMAGE004
) form, these vectors are formed matrix
Figure 189904DEST_PATH_IMAGE005
, the Gabor eigenvector in each human face region is a sample, definition between class scatter matrix
Figure 773332DEST_PATH_IMAGE006
With divergence matrix in the class For:
Figure 442714DEST_PATH_IMAGE009
Figure 513438DEST_PATH_IMAGE010
Figure 566845DEST_PATH_IMAGE011
Be the weighted value of sample, Be all
Figure 824968DEST_PATH_IMAGE013
The average of class training sample,
Figure 304359DEST_PATH_IMAGE014
Be the average of all training samples,
Figure 161457DEST_PATH_IMAGE015
Be
Figure 793427DEST_PATH_IMAGE016
In the class
Figure 444988DEST_PATH_IMAGE017
Individual training sample,
Figure 677255DEST_PATH_IMAGE018
Be the non-j sample in the i class,
Figure 72464DEST_PATH_IMAGE019
Be training sample quantity,
Figure 621257DEST_PATH_IMAGE020
Be
Figure 646982DEST_PATH_IMAGE021
The number of sample in the class,
Figure 914015DEST_PATH_IMAGE022
Be the classification sum of sample, sample
Figure 299866DEST_PATH_IMAGE023
Corresponding average is Utilize formula again
Figure 962109DEST_PATH_IMAGE025
Ask for the projection matrix of optimum improvement linear discriminant analysis
Figure 654121DEST_PATH_IMAGE026
Utilize projection matrix Training sample is carried out projection, finish the dimensionality reduction of eigenvector.
4. 1 described smiling face's recognition methods as requested is characterized in that: the support vector machine of described step (14) is two class support vector machine sorters.
CN 201110122936 2011-05-13 2011-05-13 Smiling face recognition method for reducing dimension by using improved linear discriminant analysis Expired - Fee Related CN102194107B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201110122936 CN102194107B (en) 2011-05-13 2011-05-13 Smiling face recognition method for reducing dimension by using improved linear discriminant analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201110122936 CN102194107B (en) 2011-05-13 2011-05-13 Smiling face recognition method for reducing dimension by using improved linear discriminant analysis

Publications (2)

Publication Number Publication Date
CN102194107A true CN102194107A (en) 2011-09-21
CN102194107B CN102194107B (en) 2013-03-20

Family

ID=44602151

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201110122936 Expired - Fee Related CN102194107B (en) 2011-05-13 2011-05-13 Smiling face recognition method for reducing dimension by using improved linear discriminant analysis

Country Status (1)

Country Link
CN (1) CN102194107B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103530657A (en) * 2013-09-26 2014-01-22 华南理工大学 Deep learning human face identification method based on weighting L2 extraction
CN104102903A (en) * 2014-07-14 2014-10-15 杭州电子科技大学 Secondary face recognition method based on SRC

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030086593A1 (en) * 2001-05-31 2003-05-08 Chengjun Liu Feature based classification
CN1828630A (en) * 2006-04-06 2006-09-06 上海交通大学 Manifold learning based human face posture identification method
CN1885310A (en) * 2006-06-01 2006-12-27 北京中星微电子有限公司 Human face model training module and method, human face real-time certification system and method
CN101699470A (en) * 2009-10-30 2010-04-28 华南理工大学 Extracting method for smiling face identification on picture of human face

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030086593A1 (en) * 2001-05-31 2003-05-08 Chengjun Liu Feature based classification
CN1828630A (en) * 2006-04-06 2006-09-06 上海交通大学 Manifold learning based human face posture identification method
CN1885310A (en) * 2006-06-01 2006-12-27 北京中星微电子有限公司 Human face model training module and method, human face real-time certification system and method
CN101699470A (en) * 2009-10-30 2010-04-28 华南理工大学 Extracting method for smiling face identification on picture of human face

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
叶延亮等: "基于改进的二维线性判别分析的人脸识别", 《微计算机信息(测控自动化)》 *
张玉华等: "基于线性判别分析的加权零空间算法及在人脸识别中的应用", 《山东大学学报(工学版)》 *
覃志祥等: "一种改进的线性判别分析法在人脸识别中的应用", 《计算机工程》 *
郑秋梅等: "基于双向二维直接线性判别分析的人脸表情识别", 《中国石油大学学报(自然科学版)》 *
马平等: "改进的线性判别分析及人脸识别", 《计算机与数字工程》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103530657A (en) * 2013-09-26 2014-01-22 华南理工大学 Deep learning human face identification method based on weighting L2 extraction
CN103530657B (en) * 2013-09-26 2016-07-06 华南理工大学 A kind of based on weighting L2 extraction degree of depth study face identification method
CN104102903A (en) * 2014-07-14 2014-10-15 杭州电子科技大学 Secondary face recognition method based on SRC

Also Published As

Publication number Publication date
CN102194107B (en) 2013-03-20

Similar Documents

Publication Publication Date Title
Shao et al. Deep convolutional dynamic texture learning with adaptive channel-discriminability for 3D mask face anti-spoofing
CN110070033B (en) Method for detecting wearing state of safety helmet in dangerous working area in power field
Wang et al. Hierarchical attention network for action recognition in videos
CN102194108B (en) Smile face expression recognition method based on clustering linear discriminant analysis of feature selection
CN109472198B (en) Gesture robust video smiling face recognition method
Chen et al. Survey of pedestrian action recognition techniques for autonomous driving
Ghimire et al. A robust face detection method based on skin color and edges
CN102324025A (en) Human face detection and tracking method based on Gaussian skin color model and feature analysis
Zhang et al. A survey on face anti-spoofing algorithms
CN111523462A (en) Video sequence list situation recognition system and method based on self-attention enhanced CNN
CN109753904A (en) A kind of face identification method and system
El Maghraby et al. Hybrid face detection system using combination of Viola-Jones method and skin detection
Zhu et al. Action recognition in broadcast tennis video
Yingxin et al. A robust hand gesture recognition method via convolutional neural network
Ranganatha et al. An integrated robust approach for fast face tracking in noisy real-world videos with visual constraints
Huang et al. Deepfake mnist+: a deepfake facial animation dataset
CN113920568A (en) Face and human body posture emotion recognition method based on video image
Yu et al. Pedestrian detection based on improved Faster RCNN algorithm
CN106529441A (en) Fuzzy boundary fragmentation-based depth motion map human body action recognition method
CN111666845A (en) Small sample deep learning multi-mode sign language recognition method based on key frame sampling
Chen et al. Fast face detection algorithm based on improved skin-color model
CN110008876A (en) A kind of face verification method based on data enhancing and Fusion Features
Chen et al. Fast eye detection using different color spaces
CN105893967B (en) Human behavior classification detection method and system based on time sequence retention space-time characteristics
CN102194107B (en) Smiling face recognition method for reducing dimension by using improved linear discriminant analysis

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20130320

Termination date: 20190513

CF01 Termination of patent right due to non-payment of annual fee