SlideShare a Scribd company logo
Structural Features For Face
Recognition
Prepared By: Sari Meriem
1
UNIVERSITY of MOHAMED CHERIF MESAADIA
SOUK-AHRAS
Presentation About:
Introduction
2
In this presentation we’re going to talk about features used in face
recognition and exactly geometric based features.
What are the geometric features?
Sari Meriem Structural Features 1ére Année Master GL
Introduction
3
Geometric features:
 Are features of objects constructed by a set of geometric elements like points,
lines, curves or surfaces.
 Combining machine learning and computer vision to solve visual tasks.
 Aim to find a set of representative features of a geometric form to represent an
object by collecting geometric features from images.
Sari Meriem Structural Features 1ére Année Master GL
First Article
4
First Article:
Face Recognition: Features versus Templates
By Roberto Brunelli and Tomaso Poggio
Sari Meriem Structural Features 1ére Année Master GL
Abstract
5
The principal idea:
 Go through different methods for human faces recognition by a machine.
 Chose two strategies tested on the same database to compare between them.
 Implementing one algorithme for each startegy:
 Based on geometric features with a recognition rate of 90%.
 Based on grey-level template matching with a perfect recognition rate.
Sari Meriem Structural Features 1ére Année Master GL
Used DataBase
6
 Contains 188 images.
 26 males and 21 female.
 4 pictures for each person.
 Shots from frontal view.
 Environnement exposed to sunlight through windows.
 Scale variations were about 30% max.
Sari Meriem Structural Features 1ére Année Master GL
Geometric Feature Based Matching
7
The main idea is to extract relative position and other parameters of distinctive
features such as the eyes, mouth, nose and chin.
Sari Meriem Structural Features 1ére Année Master GL
Geometric Feature Based Matching
8
 The overall configuration uses a numerical vector representing the position and
size of the main facial features.
 The set of features should satisfy the use:
 The min estimation possible.
 The min light conditions dependency possible.
 The min small facial expression change dependency.
 Get the max information possible.
 In this proposed algotithme there are three main steps:
 Normalization.
 Feature Extraction.
 Recognition Performance.
Sari Meriem Structural Features 1ére Année Master GL
Geometric Feature Based Matching
9
 Normalization:
 It allows us to get results without taking in consederation the position, scale or
rotation.
The translation dependency can be eliminated by setting the coordinates to a
point that is detected in each image.
 In this paper that point is the eye-to-eye axis.
Sari Meriem Structural Features 1ére Année Master GL
Geometric Feature Based Matching
10
 Feature Extraction:
 Face recognition offers a set of interesting constraints:
 Belatiral symetry of the face.
 Each face has two eyes, one nose and one mouth.
 It eases the task of feature extraction.
 The used extraction technique is the integral projection because it determines
the position of the features.
 Horizontal projection offers the left and right boundries of the face and the
nose.
 Vertical projection detects the head top, eyes, nose base and mouth.
Sari Meriem Structural Features 1ére Année Master GL
Geometric Feature Based Matching
11Sari Meriem Structural Features 1ére Année Master GL
Geometric Feature Based Matching
12
 In total we get 35 geometrical features extracted automaticaly in this system
and used in facial recognition as next:
 Eyebrow thickness and vertical position at the eye center position.
 A coarse description of the left eyebrow’s arches.
 Nose vertical position and width.
 Mouth vertical position, width, height upper and lower lips.
 Eleven radii describing the chin shape.
 Face width and nose position.
 Face width halfway between nose tip and eyes.
Sari Meriem Structural Features 1ére Année Master GL
Geometric Feature Based Matching
13Sari Meriem Structural Features 1ére Année Master GL
Geometric Feature Based Matching
14
 Recognition Performance:
 We already have associated previously with each face a 35 dimention numerical
vector.
 We used Bayes classifier.
 The aim of the paper is to characterize the performance as a function of the
number of the classes to be descriminated.
 We can get useful data by estemating the intraclass and the interclass variability
Find the minimum distance to a wrong correspondence over the maximum
distance to a correct correspendence.
 In this paper we represented each class by a single element so that the
maximum distance reduces the distance from the representing vector.
Sari Meriem Structural Features 1ére Année Master GL
Geometric Feature Based Matching
15Sari Meriem Structural Features 1ére Année Master GL
Template Matching
16
The simpliest version:
 Use a bidimensional array of intensity values coded image.
 Compare it with a single template representing the whole face using a metric
that is usualy euclidean distance.
 The more sophisticated versions:
 Use different viewpoints shots.
 Use a single template with qualitative prior model of how face transforms under
viewpoints change  Elastic Template Technique.
Sari Meriem Structural Features 1ére Année Master GL
Template Matching
17
 The implemented system is an extension of the work of Baron.
 We used the same normalization method cited previously.
 When finding an unknown person i.e. unclassified image, it is compared with all
of the DB images to return a vector of matching scores so that it can be classified
as the one giving the highest score.
 Used features are sorted by decreasing performance as follow:
 Eyes.
 Nose.
 Mouth.
 Whole face template.
Sari Meriem Structural Features 1ére Année Master GL
Template Matching
18Sari Meriem Structural Features 1ére Année Master GL
Template Matching
19
 Performance can be improved by using templates from more than one image
per person.
 Using templates from two images per person in the experiments has shown a
perfect recognition.
Sari Meriem Structural Features 1ére Année Master GL
Conclusion
20
 In this paper there are two developped algorithmes:
 Identification through a vector of geometrical features.
 Allow higher recognition speed.
 Smaller memory requirements(1 byte per feature).
 Identification through a template matching startegy.
 Superior recognition rate.
 Simpler.
 A successful object recognition architectures need to combine aspects of
feature-based approaches with template matching techniques.
Sari Meriem Structural Features 1ére Année Master GL
Second Article
21
Second Article:
Comparison Between Geometry-Based and Gabor-
Wavelets-Based Facial Expression Recognition Using Multi-Layer
Perceptron
Zhengyou Zhang, Michael Lyons, Michael Schuster and Shigeru Akamatsu
Sari Meriem Structural Features 1ére Année Master GL
Abstract
22
The principal idea is:
 Investigate the use of types of features in facial recognition:
 Geometric features which are a set of positions of face componenets.
 A multi-scale/oriented features called Gobor wavelet coefficients.
 The architecto developped is based on 2 layer perceptron.
 The first layer of the perceptron performs a nonlinear reduction of the dimensionality of
the feature space.
 The results set shows that gabor wavelets are more powerful than geometric features.
Sari Meriem Structural Features 1ére Année Master GL
Introduction
23
 There are a number of difficulties in FER due to the variation of facial expression,
human faces and context.
 An automatic FER system needs to detect and locate the faces, extract the facial
features classify them.
 For the face detection there lots of methods but studies showed that neural
network based ones are the most successful.
 For the feature extraction we have two approaches: geometric based and
template matching.
 In this paper facial recognition is done through elastic graph matching.
Sari Meriem Structural Features 1ére Année Master GL
Used DataBase
24
It contains 213 images of female facial expressions.
The head is almost in frontal pose.
The final distance between the eyes after rescaling is 60 pixels.
The number of images corresponding to each of the 7 categories of expression
(neutral, happiness, sadness, surprise, anger, disgust and fear) is the same.
Sari Meriem Structural Features 1ére Année Master GL
Used DataBase
25Sari Meriem Structural Features 1ére Année Master GL
Data Set and Representation
26
 Each image is represented in two ways.
 The first uses 34 facial points which
are the geometric position.
 The second uses features extracted
with 2D Gabor transforms.
Sari Meriem Structural Features 1ére Année Master GL
Architecture and Training
27
 Based on a two layer perceptron.
 Two sets of features (geometric positions and Gabor wavelet coefficients) are extracted.
 These features are fed in the input units of the two-layer perceptron.
 The first layer aims to perform a nonlinear reduction of the dimensionality of feature
space.
 The second layer makes a statistical decision based on the reduced set of features.
 An output unit is associated with a particular facial expression.
Sari Meriem Structural Features 1ére Année Master GL
Experiments
28
 We use the cross-validation technique to test different configurations of our FER
architecture.
 We have carried out experiments on the FER using the developed architecture
by using geometric positions alone, using Gabor wavelet coefficients alone, and by
using the combination of the two pieces of information.
Sari Meriem Structural Features 1ére Année Master GL
Experiments
29
 The recognition rate achieved by our system is 90.1% with 7 hidden units.
Sari Meriem Structural Features 1ére Année Master GL
Experiments After Excluding Fear Images
30
 The system achieves a generalized recognition rate of:
 73.3% with geometric positions alone.
 92.2% with Gabor wavelet coefficients alone.
 92.3% with combined information.
Sari Meriem Structural Features 1ére Année Master GL
Conclusion
31
In this paper we used of types of features in facial recognition:
 Geometric features which are a set of positions of face components.
 Recognition rate of 73.3% when used alone.
 A multi-scale/oriented features called Gobor wavelet coefficients.
 Recognition rate of 92.2% when used alone.
 A recognition rate of 92.3% when combined together.
Sari Meriem Structural Features 1ére Année Master GL

More Related Content

What's hot

IRJET- Prediction of Facial Attribute without Landmark Information
IRJET-  	  Prediction of Facial Attribute without Landmark InformationIRJET-  	  Prediction of Facial Attribute without Landmark Information
IRJET- Prediction of Facial Attribute without Landmark Information
IRJET Journal
 
FACIAL EXPRESSION RECOGNITION USING DIGITALISED FACIAL FEATURES BASED ON ACTI...
FACIAL EXPRESSION RECOGNITION USING DIGITALISED FACIAL FEATURES BASED ON ACTI...FACIAL EXPRESSION RECOGNITION USING DIGITALISED FACIAL FEATURES BASED ON ACTI...
FACIAL EXPRESSION RECOGNITION USING DIGITALISED FACIAL FEATURES BASED ON ACTI...
csandit
 
20320130406016
2032013040601620320130406016
20320130406016
IAEME Publication
 
Facial expression identification by using features of salient facial landmarks
Facial expression identification by using features of salient facial landmarksFacial expression identification by using features of salient facial landmarks
Facial expression identification by using features of salient facial landmarks
eSAT Journals
 
Facial expression recognition based on image feature
Facial expression recognition based on image featureFacial expression recognition based on image feature
Facial expression recognition based on image feature
Tasnim Tara
 
EFFECT OF FACE TAMPERING ON FACE RECOGNITION
EFFECT OF FACE TAMPERING ON FACE RECOGNITIONEFFECT OF FACE TAMPERING ON FACE RECOGNITION
EFFECT OF FACE TAMPERING ON FACE RECOGNITION
sipij
 
Facial expression recognition on real world face images (OPTIK)
Facial expression recognition on real world face images (OPTIK)Facial expression recognition on real world face images (OPTIK)
Facial expression recognition on real world face images (OPTIK)
Sohail Ahmed
 
Ijarcet vol-2-issue-4-1352-1356
Ijarcet vol-2-issue-4-1352-1356Ijarcet vol-2-issue-4-1352-1356
Ijarcet vol-2-issue-4-1352-1356
Editor IJARCET
 
A study on face recognition technique based on eigenface
A study on face recognition technique based on eigenfaceA study on face recognition technique based on eigenface
A study on face recognition technique based on eigenface
sadique_ghitm
 
An Accurate Facial Component Detection Using Gabor Filter
An Accurate Facial Component Detection Using Gabor FilterAn Accurate Facial Component Detection Using Gabor Filter
An Accurate Facial Component Detection Using Gabor Filter
journalBEEI
 
IRJET- Face Spoof Detection using Machine Learning with Colour Features
IRJET-  	  Face Spoof Detection using Machine Learning with Colour FeaturesIRJET-  	  Face Spoof Detection using Machine Learning with Colour Features
IRJET- Face Spoof Detection using Machine Learning with Colour Features
IRJET Journal
 
WCTFR : W RAPPING C URVELET T RANSFORM B ASED F ACE R ECOGNITION
WCTFR : W RAPPING  C URVELET T RANSFORM  B ASED  F ACE  R ECOGNITIONWCTFR : W RAPPING  C URVELET T RANSFORM  B ASED  F ACE  R ECOGNITION
WCTFR : W RAPPING C URVELET T RANSFORM B ASED F ACE R ECOGNITION
csandit
 
Semantic 3DTV Content Analysis and Description
Semantic 3DTV Content Analysis and DescriptionSemantic 3DTV Content Analysis and Description
Semantic 3DTV Content Analysis and Description
Distinguished Lecturer Series - Leon The Mathematician
 
Detection and rectification of distorted fingerprints
Detection and rectification of distorted fingerprintsDetection and rectification of distorted fingerprints
Detection and rectification of distorted fingerprints
I3E Technologies
 
Image processing
Image processingImage processing
Image processing
Vinod Gurram
 
Detection and rectification of distorted fingerprints
Detection and rectification of distorted fingerprintsDetection and rectification of distorted fingerprints
Detection and rectification of distorted fingerprints
Pvrtechnologies Nellore
 
Recognition of Partially Occluded Face Using Gradientface and Local Binary Pa...
Recognition of Partially Occluded Face Using Gradientface and Local Binary Pa...Recognition of Partially Occluded Face Using Gradientface and Local Binary Pa...
Recognition of Partially Occluded Face Using Gradientface and Local Binary Pa...
Win Yu
 
3D Dynamic Facial Sequences Analsysis for face recognition and emotion detection
3D Dynamic Facial Sequences Analsysis for face recognition and emotion detection3D Dynamic Facial Sequences Analsysis for face recognition and emotion detection
3D Dynamic Facial Sequences Analsysis for face recognition and emotion detection
Taleb ALASHKAR
 
IRJET- A Review on Face Recognition using Local Binary Pattern Algorithm
IRJET- A Review on Face Recognition using Local Binary Pattern AlgorithmIRJET- A Review on Face Recognition using Local Binary Pattern Algorithm
IRJET- A Review on Face Recognition using Local Binary Pattern Algorithm
IRJET Journal
 
Ch 2
Ch 2Ch 2

What's hot (20)

IRJET- Prediction of Facial Attribute without Landmark Information
IRJET-  	  Prediction of Facial Attribute without Landmark InformationIRJET-  	  Prediction of Facial Attribute without Landmark Information
IRJET- Prediction of Facial Attribute without Landmark Information
 
FACIAL EXPRESSION RECOGNITION USING DIGITALISED FACIAL FEATURES BASED ON ACTI...
FACIAL EXPRESSION RECOGNITION USING DIGITALISED FACIAL FEATURES BASED ON ACTI...FACIAL EXPRESSION RECOGNITION USING DIGITALISED FACIAL FEATURES BASED ON ACTI...
FACIAL EXPRESSION RECOGNITION USING DIGITALISED FACIAL FEATURES BASED ON ACTI...
 
20320130406016
2032013040601620320130406016
20320130406016
 
Facial expression identification by using features of salient facial landmarks
Facial expression identification by using features of salient facial landmarksFacial expression identification by using features of salient facial landmarks
Facial expression identification by using features of salient facial landmarks
 
Facial expression recognition based on image feature
Facial expression recognition based on image featureFacial expression recognition based on image feature
Facial expression recognition based on image feature
 
EFFECT OF FACE TAMPERING ON FACE RECOGNITION
EFFECT OF FACE TAMPERING ON FACE RECOGNITIONEFFECT OF FACE TAMPERING ON FACE RECOGNITION
EFFECT OF FACE TAMPERING ON FACE RECOGNITION
 
Facial expression recognition on real world face images (OPTIK)
Facial expression recognition on real world face images (OPTIK)Facial expression recognition on real world face images (OPTIK)
Facial expression recognition on real world face images (OPTIK)
 
Ijarcet vol-2-issue-4-1352-1356
Ijarcet vol-2-issue-4-1352-1356Ijarcet vol-2-issue-4-1352-1356
Ijarcet vol-2-issue-4-1352-1356
 
A study on face recognition technique based on eigenface
A study on face recognition technique based on eigenfaceA study on face recognition technique based on eigenface
A study on face recognition technique based on eigenface
 
An Accurate Facial Component Detection Using Gabor Filter
An Accurate Facial Component Detection Using Gabor FilterAn Accurate Facial Component Detection Using Gabor Filter
An Accurate Facial Component Detection Using Gabor Filter
 
IRJET- Face Spoof Detection using Machine Learning with Colour Features
IRJET-  	  Face Spoof Detection using Machine Learning with Colour FeaturesIRJET-  	  Face Spoof Detection using Machine Learning with Colour Features
IRJET- Face Spoof Detection using Machine Learning with Colour Features
 
WCTFR : W RAPPING C URVELET T RANSFORM B ASED F ACE R ECOGNITION
WCTFR : W RAPPING  C URVELET T RANSFORM  B ASED  F ACE  R ECOGNITIONWCTFR : W RAPPING  C URVELET T RANSFORM  B ASED  F ACE  R ECOGNITION
WCTFR : W RAPPING C URVELET T RANSFORM B ASED F ACE R ECOGNITION
 
Semantic 3DTV Content Analysis and Description
Semantic 3DTV Content Analysis and DescriptionSemantic 3DTV Content Analysis and Description
Semantic 3DTV Content Analysis and Description
 
Detection and rectification of distorted fingerprints
Detection and rectification of distorted fingerprintsDetection and rectification of distorted fingerprints
Detection and rectification of distorted fingerprints
 
Image processing
Image processingImage processing
Image processing
 
Detection and rectification of distorted fingerprints
Detection and rectification of distorted fingerprintsDetection and rectification of distorted fingerprints
Detection and rectification of distorted fingerprints
 
Recognition of Partially Occluded Face Using Gradientface and Local Binary Pa...
Recognition of Partially Occluded Face Using Gradientface and Local Binary Pa...Recognition of Partially Occluded Face Using Gradientface and Local Binary Pa...
Recognition of Partially Occluded Face Using Gradientface and Local Binary Pa...
 
3D Dynamic Facial Sequences Analsysis for face recognition and emotion detection
3D Dynamic Facial Sequences Analsysis for face recognition and emotion detection3D Dynamic Facial Sequences Analsysis for face recognition and emotion detection
3D Dynamic Facial Sequences Analsysis for face recognition and emotion detection
 
IRJET- A Review on Face Recognition using Local Binary Pattern Algorithm
IRJET- A Review on Face Recognition using Local Binary Pattern AlgorithmIRJET- A Review on Face Recognition using Local Binary Pattern Algorithm
IRJET- A Review on Face Recognition using Local Binary Pattern Algorithm
 
Ch 2
Ch 2Ch 2
Ch 2
 

Viewers also liked

Rest ws
Rest wsRest ws
Rest ws
meriem sari
 
Cvs
CvsCvs
Contraintes uml
Contraintes umlContraintes uml
Contraintes uml
meriem sari
 
Diagrammes de classes uml
Diagrammes de classes umlDiagrammes de classes uml
Diagrammes de classes uml
meriem sari
 
Ws asp.net
Ws asp.netWs asp.net
Ws asp.net
meriem sari
 
Meta-modelisation avec eclipse
Meta-modelisation avec eclipseMeta-modelisation avec eclipse
Meta-modelisation avec eclipse
meriem sari
 
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtoseconde
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser FemtosecondeMerieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtoseconde
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtoseconde
Yassin Sabar
 
Re l ex smile, the future of refractive surgery - medifocus sept 2013
Re l ex smile, the future of refractive surgery - medifocus sept 2013Re l ex smile, the future of refractive surgery - medifocus sept 2013
Re l ex smile, the future of refractive surgery - medifocus sept 2013
Medifocus
 
Relex Smile laser surgery
Relex Smile laser surgeryRelex Smile laser surgery
Relex Smile laser surgery
Frank Jr Goes
 
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtosec...
Merieme HAROUCHE : Le Relex Smile Correction de la myopie  au laser Femtosec...Merieme HAROUCHE : Le Relex Smile Correction de la myopie  au laser Femtosec...
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtosec...
Yassin Sabar
 
La & edm in practice
La & edm in practiceLa & edm in practice
La & edm in practice
bharati k
 
Ressource numérique Circuit électrique au primaire
Ressource numérique Circuit électrique au primaire Ressource numérique Circuit électrique au primaire
Ressource numérique Circuit électrique au primaire
Erradi Mohamed
 
UMLX and QVT and ATL
UMLX and QVT and ATLUMLX and QVT and ATL
UMLX and QVT and ATL
Edward Willink
 
OCCIware Contribution to the EU consultation on Cloud Computing Research Inno...
OCCIware Contribution to the EU consultation on Cloud Computing Research Inno...OCCIware Contribution to the EU consultation on Cloud Computing Research Inno...
OCCIware Contribution to the EU consultation on Cloud Computing Research Inno...
OCCIware
 
Collaboration and Governance of Open Source Projects
Collaboration and Governance of Open Source ProjectsCollaboration and Governance of Open Source Projects
Collaboration and Governance of Open Source Projects
Jordi Cabot
 
Cvl
CvlCvl
Fast, Faster and Super-Fast Queries
Fast, Faster and Super-Fast QueriesFast, Faster and Super-Fast Queries
Fast, Faster and Super-Fast Queries
Edward Willink
 
SysML adoption in France
SysML adoption in FranceSysML adoption in France
SysML adoption in FrancePascal Roques
 
Vbisigk
VbisigkVbisigk
VbisigkISIG
 
Design Thinking Assignment
Design Thinking AssignmentDesign Thinking Assignment
Design Thinking Assignment
Salma ES-Salmani
 

Viewers also liked (20)

Rest ws
Rest wsRest ws
Rest ws
 
Cvs
CvsCvs
Cvs
 
Contraintes uml
Contraintes umlContraintes uml
Contraintes uml
 
Diagrammes de classes uml
Diagrammes de classes umlDiagrammes de classes uml
Diagrammes de classes uml
 
Ws asp.net
Ws asp.netWs asp.net
Ws asp.net
 
Meta-modelisation avec eclipse
Meta-modelisation avec eclipseMeta-modelisation avec eclipse
Meta-modelisation avec eclipse
 
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtoseconde
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser FemtosecondeMerieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtoseconde
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtoseconde
 
Re l ex smile, the future of refractive surgery - medifocus sept 2013
Re l ex smile, the future of refractive surgery - medifocus sept 2013Re l ex smile, the future of refractive surgery - medifocus sept 2013
Re l ex smile, the future of refractive surgery - medifocus sept 2013
 
Relex Smile laser surgery
Relex Smile laser surgeryRelex Smile laser surgery
Relex Smile laser surgery
 
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtosec...
Merieme HAROUCHE : Le Relex Smile Correction de la myopie  au laser Femtosec...Merieme HAROUCHE : Le Relex Smile Correction de la myopie  au laser Femtosec...
Merieme HAROUCHE : Le Relex Smile Correction de la myopie au laser Femtosec...
 
La & edm in practice
La & edm in practiceLa & edm in practice
La & edm in practice
 
Ressource numérique Circuit électrique au primaire
Ressource numérique Circuit électrique au primaire Ressource numérique Circuit électrique au primaire
Ressource numérique Circuit électrique au primaire
 
UMLX and QVT and ATL
UMLX and QVT and ATLUMLX and QVT and ATL
UMLX and QVT and ATL
 
OCCIware Contribution to the EU consultation on Cloud Computing Research Inno...
OCCIware Contribution to the EU consultation on Cloud Computing Research Inno...OCCIware Contribution to the EU consultation on Cloud Computing Research Inno...
OCCIware Contribution to the EU consultation on Cloud Computing Research Inno...
 
Collaboration and Governance of Open Source Projects
Collaboration and Governance of Open Source ProjectsCollaboration and Governance of Open Source Projects
Collaboration and Governance of Open Source Projects
 
Cvl
CvlCvl
Cvl
 
Fast, Faster and Super-Fast Queries
Fast, Faster and Super-Fast QueriesFast, Faster and Super-Fast Queries
Fast, Faster and Super-Fast Queries
 
SysML adoption in France
SysML adoption in FranceSysML adoption in France
SysML adoption in France
 
Vbisigk
VbisigkVbisigk
Vbisigk
 
Design Thinking Assignment
Design Thinking AssignmentDesign Thinking Assignment
Design Thinking Assignment
 

Similar to Structural features for face recognition

A novel approach for performance parameter estimation of face recognition bas...
A novel approach for performance parameter estimation of face recognition bas...A novel approach for performance parameter estimation of face recognition bas...
A novel approach for performance parameter estimation of face recognition bas...
IJMER
 
Model Based Emotion Detection using Point Clouds
Model Based Emotion Detection using Point CloudsModel Based Emotion Detection using Point Clouds
Model Based Emotion Detection using Point Clouds
Lakshmi Sarvani Videla
 
Face recognition
Face recognitionFace recognition
Face recognition
Satyendra Rajput
 
Real time facial expression analysis using pca
Real time facial expression analysis using pcaReal time facial expression analysis using pca
Real time facial expression analysis using pca
International Journal of Science and Research (IJSR)
 
Ck36515520
Ck36515520Ck36515520
Ck36515520
IJERA Editor
 
Lc3420022006
Lc3420022006Lc3420022006
Lc3420022006
IJERA Editor
 
Facial expression identification by using features of salient facial landmarks
Facial expression identification by using features of salient facial landmarksFacial expression identification by using features of salient facial landmarks
Facial expression identification by using features of salient facial landmarks
eSAT Journals
 
Critical evaluation of frontal image based gender classification techniques
Critical evaluation of frontal image based gender classification techniquesCritical evaluation of frontal image based gender classification techniques
Critical evaluation of frontal image based gender classification techniques
Salam Shah
 
face recognition system
face recognition systemface recognition system
face recognition system
Anil Kumar
 
Ijetcas14 351
Ijetcas14 351Ijetcas14 351
Ijetcas14 351
Iasir Journals
 
Aa4102207210
Aa4102207210Aa4102207210
Aa4102207210
IJERA Editor
 
A study of techniques for facial detection and expression classification
A study of techniques for facial detection and expression classificationA study of techniques for facial detection and expression classification
A study of techniques for facial detection and expression classification
IJCSES Journal
 
A Novel Mathematical Based Method for Generating Virtual Samples from a Front...
A Novel Mathematical Based Method for Generating Virtual Samples from a Front...A Novel Mathematical Based Method for Generating Virtual Samples from a Front...
A Novel Mathematical Based Method for Generating Virtual Samples from a Front...
CSCJournals
 
IRJET- Class Attendance using Face Detection and Recognition with OPENCV
IRJET- Class Attendance using Face Detection and Recognition with OPENCVIRJET- Class Attendance using Face Detection and Recognition with OPENCV
IRJET- Class Attendance using Face Detection and Recognition with OPENCV
IRJET Journal
 
IRJET- A Review on Various Approaches of Face Recognition
IRJET- A Review on Various Approaches of Face RecognitionIRJET- A Review on Various Approaches of Face Recognition
IRJET- A Review on Various Approaches of Face Recognition
IRJET Journal
 
Face Recognition Using Gabor features And PCA
Face Recognition Using Gabor features And PCAFace Recognition Using Gabor features And PCA
Face Recognition Using Gabor features And PCA
IOSR Journals
 
IRJET - Facial Recognition based Attendance System with LBPH
IRJET -  	  Facial Recognition based Attendance System with LBPHIRJET -  	  Facial Recognition based Attendance System with LBPH
IRJET - Facial Recognition based Attendance System with LBPH
IRJET Journal
 
A017530114
A017530114A017530114
A017530114
IOSR Journals
 
Eigenfaces , Fisherfaces and Dimensionality_Reduction
Eigenfaces , Fisherfaces and Dimensionality_ReductionEigenfaces , Fisherfaces and Dimensionality_Reduction
Eigenfaces , Fisherfaces and Dimensionality_Reduction
mostafayounes012
 
International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)
IJERD Editor
 

Similar to Structural features for face recognition (20)

A novel approach for performance parameter estimation of face recognition bas...
A novel approach for performance parameter estimation of face recognition bas...A novel approach for performance parameter estimation of face recognition bas...
A novel approach for performance parameter estimation of face recognition bas...
 
Model Based Emotion Detection using Point Clouds
Model Based Emotion Detection using Point CloudsModel Based Emotion Detection using Point Clouds
Model Based Emotion Detection using Point Clouds
 
Face recognition
Face recognitionFace recognition
Face recognition
 
Real time facial expression analysis using pca
Real time facial expression analysis using pcaReal time facial expression analysis using pca
Real time facial expression analysis using pca
 
Ck36515520
Ck36515520Ck36515520
Ck36515520
 
Lc3420022006
Lc3420022006Lc3420022006
Lc3420022006
 
Facial expression identification by using features of salient facial landmarks
Facial expression identification by using features of salient facial landmarksFacial expression identification by using features of salient facial landmarks
Facial expression identification by using features of salient facial landmarks
 
Critical evaluation of frontal image based gender classification techniques
Critical evaluation of frontal image based gender classification techniquesCritical evaluation of frontal image based gender classification techniques
Critical evaluation of frontal image based gender classification techniques
 
face recognition system
face recognition systemface recognition system
face recognition system
 
Ijetcas14 351
Ijetcas14 351Ijetcas14 351
Ijetcas14 351
 
Aa4102207210
Aa4102207210Aa4102207210
Aa4102207210
 
A study of techniques for facial detection and expression classification
A study of techniques for facial detection and expression classificationA study of techniques for facial detection and expression classification
A study of techniques for facial detection and expression classification
 
A Novel Mathematical Based Method for Generating Virtual Samples from a Front...
A Novel Mathematical Based Method for Generating Virtual Samples from a Front...A Novel Mathematical Based Method for Generating Virtual Samples from a Front...
A Novel Mathematical Based Method for Generating Virtual Samples from a Front...
 
IRJET- Class Attendance using Face Detection and Recognition with OPENCV
IRJET- Class Attendance using Face Detection and Recognition with OPENCVIRJET- Class Attendance using Face Detection and Recognition with OPENCV
IRJET- Class Attendance using Face Detection and Recognition with OPENCV
 
IRJET- A Review on Various Approaches of Face Recognition
IRJET- A Review on Various Approaches of Face RecognitionIRJET- A Review on Various Approaches of Face Recognition
IRJET- A Review on Various Approaches of Face Recognition
 
Face Recognition Using Gabor features And PCA
Face Recognition Using Gabor features And PCAFace Recognition Using Gabor features And PCA
Face Recognition Using Gabor features And PCA
 
IRJET - Facial Recognition based Attendance System with LBPH
IRJET -  	  Facial Recognition based Attendance System with LBPHIRJET -  	  Facial Recognition based Attendance System with LBPH
IRJET - Facial Recognition based Attendance System with LBPH
 
A017530114
A017530114A017530114
A017530114
 
Eigenfaces , Fisherfaces and Dimensionality_Reduction
Eigenfaces , Fisherfaces and Dimensionality_ReductionEigenfaces , Fisherfaces and Dimensionality_Reduction
Eigenfaces , Fisherfaces and Dimensionality_Reduction
 
International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)
 

Recently uploaded

Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging DataPredictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
Kiwi Creative
 
A presentation that explain the Power BI Licensing
A presentation that explain the Power BI LicensingA presentation that explain the Power BI Licensing
A presentation that explain the Power BI Licensing
AlessioFois2
 
一比一原版(曼大毕业证书)曼尼托巴大学毕业证如何办理
一比一原版(曼大毕业证书)曼尼托巴大学毕业证如何办理一比一原版(曼大毕业证书)曼尼托巴大学毕业证如何办理
一比一原版(曼大毕业证书)曼尼托巴大学毕业证如何办理
ytypuem
 
[VCOSA] Monthly Report - Cotton & Yarn Statistics March 2024
[VCOSA] Monthly Report - Cotton & Yarn Statistics March 2024[VCOSA] Monthly Report - Cotton & Yarn Statistics March 2024
[VCOSA] Monthly Report - Cotton & Yarn Statistics March 2024
Vietnam Cotton & Spinning Association
 
Sample Devops SRE Product Companies .pdf
Sample Devops SRE  Product Companies .pdfSample Devops SRE  Product Companies .pdf
Sample Devops SRE Product Companies .pdf
Vineet
 
University of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma TranscriptUniversity of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma Transcript
soxrziqu
 
Beyond the Basics of A/B Tests: Highly Innovative Experimentation Tactics You...
Beyond the Basics of A/B Tests: Highly Innovative Experimentation Tactics You...Beyond the Basics of A/B Tests: Highly Innovative Experimentation Tactics You...
Beyond the Basics of A/B Tests: Highly Innovative Experimentation Tactics You...
Aggregage
 
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
nuttdpt
 
一比一原版卡尔加里大学毕业证(uc毕业证)如何办理
一比一原版卡尔加里大学毕业证(uc毕业证)如何办理一比一原版卡尔加里大学毕业证(uc毕业证)如何办理
一比一原版卡尔加里大学毕业证(uc毕业证)如何办理
oaxefes
 
Open Source Contributions to Postgres: The Basics POSETTE 2024
Open Source Contributions to Postgres: The Basics POSETTE 2024Open Source Contributions to Postgres: The Basics POSETTE 2024
Open Source Contributions to Postgres: The Basics POSETTE 2024
ElizabethGarrettChri
 
一比一原版英属哥伦比亚大学毕业证(UBC毕业证书)学历如何办理
一比一原版英属哥伦比亚大学毕业证(UBC毕业证书)学历如何办理一比一原版英属哥伦比亚大学毕业证(UBC毕业证书)学历如何办理
一比一原版英属哥伦比亚大学毕业证(UBC毕业证书)学历如何办理
z6osjkqvd
 
DSSML24_tspann_CodelessGenerativeAIPipelines
DSSML24_tspann_CodelessGenerativeAIPipelinesDSSML24_tspann_CodelessGenerativeAIPipelines
DSSML24_tspann_CodelessGenerativeAIPipelines
Timothy Spann
 
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
mkkikqvo
 
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
nuttdpt
 
一比一原版(Unimelb毕业证书)墨尔本大学毕业证如何办理
一比一原版(Unimelb毕业证书)墨尔本大学毕业证如何办理一比一原版(Unimelb毕业证书)墨尔本大学毕业证如何办理
一比一原版(Unimelb毕业证书)墨尔本大学毕业证如何办理
xclpvhuk
 
一比一原版南十字星大学毕业证(SCU毕业证书)学历如何办理
一比一原版南十字星大学毕业证(SCU毕业证书)学历如何办理一比一原版南十字星大学毕业证(SCU毕业证书)学历如何办理
一比一原版南十字星大学毕业证(SCU毕业证书)学历如何办理
slg6lamcq
 
Experts live - Improving user adoption with AI
Experts live - Improving user adoption with AIExperts live - Improving user adoption with AI
Experts live - Improving user adoption with AI
jitskeb
 
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
nyfuhyz
 
一比一原版爱尔兰都柏林大学毕业证(本硕)ucd学位证书如何办理
一比一原版爱尔兰都柏林大学毕业证(本硕)ucd学位证书如何办理一比一原版爱尔兰都柏林大学毕业证(本硕)ucd学位证书如何办理
一比一原版爱尔兰都柏林大学毕业证(本硕)ucd学位证书如何办理
hqfek
 
writing report business partner b1+ .pdf
writing report business partner b1+ .pdfwriting report business partner b1+ .pdf
writing report business partner b1+ .pdf
VyNguyen709676
 

Recently uploaded (20)

Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging DataPredictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
 
A presentation that explain the Power BI Licensing
A presentation that explain the Power BI LicensingA presentation that explain the Power BI Licensing
A presentation that explain the Power BI Licensing
 
一比一原版(曼大毕业证书)曼尼托巴大学毕业证如何办理
一比一原版(曼大毕业证书)曼尼托巴大学毕业证如何办理一比一原版(曼大毕业证书)曼尼托巴大学毕业证如何办理
一比一原版(曼大毕业证书)曼尼托巴大学毕业证如何办理
 
[VCOSA] Monthly Report - Cotton & Yarn Statistics March 2024
[VCOSA] Monthly Report - Cotton & Yarn Statistics March 2024[VCOSA] Monthly Report - Cotton & Yarn Statistics March 2024
[VCOSA] Monthly Report - Cotton & Yarn Statistics March 2024
 
Sample Devops SRE Product Companies .pdf
Sample Devops SRE  Product Companies .pdfSample Devops SRE  Product Companies .pdf
Sample Devops SRE Product Companies .pdf
 
University of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma TranscriptUniversity of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma Transcript
 
Beyond the Basics of A/B Tests: Highly Innovative Experimentation Tactics You...
Beyond the Basics of A/B Tests: Highly Innovative Experimentation Tactics You...Beyond the Basics of A/B Tests: Highly Innovative Experimentation Tactics You...
Beyond the Basics of A/B Tests: Highly Innovative Experimentation Tactics You...
 
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
一比一原版(UCSF文凭证书)旧金山分校毕业证如何办理
 
一比一原版卡尔加里大学毕业证(uc毕业证)如何办理
一比一原版卡尔加里大学毕业证(uc毕业证)如何办理一比一原版卡尔加里大学毕业证(uc毕业证)如何办理
一比一原版卡尔加里大学毕业证(uc毕业证)如何办理
 
Open Source Contributions to Postgres: The Basics POSETTE 2024
Open Source Contributions to Postgres: The Basics POSETTE 2024Open Source Contributions to Postgres: The Basics POSETTE 2024
Open Source Contributions to Postgres: The Basics POSETTE 2024
 
一比一原版英属哥伦比亚大学毕业证(UBC毕业证书)学历如何办理
一比一原版英属哥伦比亚大学毕业证(UBC毕业证书)学历如何办理一比一原版英属哥伦比亚大学毕业证(UBC毕业证书)学历如何办理
一比一原版英属哥伦比亚大学毕业证(UBC毕业证书)学历如何办理
 
DSSML24_tspann_CodelessGenerativeAIPipelines
DSSML24_tspann_CodelessGenerativeAIPipelinesDSSML24_tspann_CodelessGenerativeAIPipelines
DSSML24_tspann_CodelessGenerativeAIPipelines
 
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
原版一比一多伦多大学毕业证(UofT毕业证书)如何办理
 
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
 
一比一原版(Unimelb毕业证书)墨尔本大学毕业证如何办理
一比一原版(Unimelb毕业证书)墨尔本大学毕业证如何办理一比一原版(Unimelb毕业证书)墨尔本大学毕业证如何办理
一比一原版(Unimelb毕业证书)墨尔本大学毕业证如何办理
 
一比一原版南十字星大学毕业证(SCU毕业证书)学历如何办理
一比一原版南十字星大学毕业证(SCU毕业证书)学历如何办理一比一原版南十字星大学毕业证(SCU毕业证书)学历如何办理
一比一原版南十字星大学毕业证(SCU毕业证书)学历如何办理
 
Experts live - Improving user adoption with AI
Experts live - Improving user adoption with AIExperts live - Improving user adoption with AI
Experts live - Improving user adoption with AI
 
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
 
一比一原版爱尔兰都柏林大学毕业证(本硕)ucd学位证书如何办理
一比一原版爱尔兰都柏林大学毕业证(本硕)ucd学位证书如何办理一比一原版爱尔兰都柏林大学毕业证(本硕)ucd学位证书如何办理
一比一原版爱尔兰都柏林大学毕业证(本硕)ucd学位证书如何办理
 
writing report business partner b1+ .pdf
writing report business partner b1+ .pdfwriting report business partner b1+ .pdf
writing report business partner b1+ .pdf
 

Structural features for face recognition

  • 1. Structural Features For Face Recognition Prepared By: Sari Meriem 1 UNIVERSITY of MOHAMED CHERIF MESAADIA SOUK-AHRAS Presentation About:
  • 2. Introduction 2 In this presentation we’re going to talk about features used in face recognition and exactly geometric based features. What are the geometric features? Sari Meriem Structural Features 1ére Année Master GL
  • 3. Introduction 3 Geometric features:  Are features of objects constructed by a set of geometric elements like points, lines, curves or surfaces.  Combining machine learning and computer vision to solve visual tasks.  Aim to find a set of representative features of a geometric form to represent an object by collecting geometric features from images. Sari Meriem Structural Features 1ére Année Master GL
  • 4. First Article 4 First Article: Face Recognition: Features versus Templates By Roberto Brunelli and Tomaso Poggio Sari Meriem Structural Features 1ére Année Master GL
  • 5. Abstract 5 The principal idea:  Go through different methods for human faces recognition by a machine.  Chose two strategies tested on the same database to compare between them.  Implementing one algorithme for each startegy:  Based on geometric features with a recognition rate of 90%.  Based on grey-level template matching with a perfect recognition rate. Sari Meriem Structural Features 1ére Année Master GL
  • 6. Used DataBase 6  Contains 188 images.  26 males and 21 female.  4 pictures for each person.  Shots from frontal view.  Environnement exposed to sunlight through windows.  Scale variations were about 30% max. Sari Meriem Structural Features 1ére Année Master GL
  • 7. Geometric Feature Based Matching 7 The main idea is to extract relative position and other parameters of distinctive features such as the eyes, mouth, nose and chin. Sari Meriem Structural Features 1ére Année Master GL
  • 8. Geometric Feature Based Matching 8  The overall configuration uses a numerical vector representing the position and size of the main facial features.  The set of features should satisfy the use:  The min estimation possible.  The min light conditions dependency possible.  The min small facial expression change dependency.  Get the max information possible.  In this proposed algotithme there are three main steps:  Normalization.  Feature Extraction.  Recognition Performance. Sari Meriem Structural Features 1ére Année Master GL
  • 9. Geometric Feature Based Matching 9  Normalization:  It allows us to get results without taking in consederation the position, scale or rotation. The translation dependency can be eliminated by setting the coordinates to a point that is detected in each image.  In this paper that point is the eye-to-eye axis. Sari Meriem Structural Features 1ére Année Master GL
  • 10. Geometric Feature Based Matching 10  Feature Extraction:  Face recognition offers a set of interesting constraints:  Belatiral symetry of the face.  Each face has two eyes, one nose and one mouth.  It eases the task of feature extraction.  The used extraction technique is the integral projection because it determines the position of the features.  Horizontal projection offers the left and right boundries of the face and the nose.  Vertical projection detects the head top, eyes, nose base and mouth. Sari Meriem Structural Features 1ére Année Master GL
  • 11. Geometric Feature Based Matching 11Sari Meriem Structural Features 1ére Année Master GL
  • 12. Geometric Feature Based Matching 12  In total we get 35 geometrical features extracted automaticaly in this system and used in facial recognition as next:  Eyebrow thickness and vertical position at the eye center position.  A coarse description of the left eyebrow’s arches.  Nose vertical position and width.  Mouth vertical position, width, height upper and lower lips.  Eleven radii describing the chin shape.  Face width and nose position.  Face width halfway between nose tip and eyes. Sari Meriem Structural Features 1ére Année Master GL
  • 13. Geometric Feature Based Matching 13Sari Meriem Structural Features 1ére Année Master GL
  • 14. Geometric Feature Based Matching 14  Recognition Performance:  We already have associated previously with each face a 35 dimention numerical vector.  We used Bayes classifier.  The aim of the paper is to characterize the performance as a function of the number of the classes to be descriminated.  We can get useful data by estemating the intraclass and the interclass variability Find the minimum distance to a wrong correspondence over the maximum distance to a correct correspendence.  In this paper we represented each class by a single element so that the maximum distance reduces the distance from the representing vector. Sari Meriem Structural Features 1ére Année Master GL
  • 15. Geometric Feature Based Matching 15Sari Meriem Structural Features 1ére Année Master GL
  • 16. Template Matching 16 The simpliest version:  Use a bidimensional array of intensity values coded image.  Compare it with a single template representing the whole face using a metric that is usualy euclidean distance.  The more sophisticated versions:  Use different viewpoints shots.  Use a single template with qualitative prior model of how face transforms under viewpoints change  Elastic Template Technique. Sari Meriem Structural Features 1ére Année Master GL
  • 17. Template Matching 17  The implemented system is an extension of the work of Baron.  We used the same normalization method cited previously.  When finding an unknown person i.e. unclassified image, it is compared with all of the DB images to return a vector of matching scores so that it can be classified as the one giving the highest score.  Used features are sorted by decreasing performance as follow:  Eyes.  Nose.  Mouth.  Whole face template. Sari Meriem Structural Features 1ére Année Master GL
  • 18. Template Matching 18Sari Meriem Structural Features 1ére Année Master GL
  • 19. Template Matching 19  Performance can be improved by using templates from more than one image per person.  Using templates from two images per person in the experiments has shown a perfect recognition. Sari Meriem Structural Features 1ére Année Master GL
  • 20. Conclusion 20  In this paper there are two developped algorithmes:  Identification through a vector of geometrical features.  Allow higher recognition speed.  Smaller memory requirements(1 byte per feature).  Identification through a template matching startegy.  Superior recognition rate.  Simpler.  A successful object recognition architectures need to combine aspects of feature-based approaches with template matching techniques. Sari Meriem Structural Features 1ére Année Master GL
  • 21. Second Article 21 Second Article: Comparison Between Geometry-Based and Gabor- Wavelets-Based Facial Expression Recognition Using Multi-Layer Perceptron Zhengyou Zhang, Michael Lyons, Michael Schuster and Shigeru Akamatsu Sari Meriem Structural Features 1ére Année Master GL
  • 22. Abstract 22 The principal idea is:  Investigate the use of types of features in facial recognition:  Geometric features which are a set of positions of face componenets.  A multi-scale/oriented features called Gobor wavelet coefficients.  The architecto developped is based on 2 layer perceptron.  The first layer of the perceptron performs a nonlinear reduction of the dimensionality of the feature space.  The results set shows that gabor wavelets are more powerful than geometric features. Sari Meriem Structural Features 1ére Année Master GL
  • 23. Introduction 23  There are a number of difficulties in FER due to the variation of facial expression, human faces and context.  An automatic FER system needs to detect and locate the faces, extract the facial features classify them.  For the face detection there lots of methods but studies showed that neural network based ones are the most successful.  For the feature extraction we have two approaches: geometric based and template matching.  In this paper facial recognition is done through elastic graph matching. Sari Meriem Structural Features 1ére Année Master GL
  • 24. Used DataBase 24 It contains 213 images of female facial expressions. The head is almost in frontal pose. The final distance between the eyes after rescaling is 60 pixels. The number of images corresponding to each of the 7 categories of expression (neutral, happiness, sadness, surprise, anger, disgust and fear) is the same. Sari Meriem Structural Features 1ére Année Master GL
  • 25. Used DataBase 25Sari Meriem Structural Features 1ére Année Master GL
  • 26. Data Set and Representation 26  Each image is represented in two ways.  The first uses 34 facial points which are the geometric position.  The second uses features extracted with 2D Gabor transforms. Sari Meriem Structural Features 1ére Année Master GL
  • 27. Architecture and Training 27  Based on a two layer perceptron.  Two sets of features (geometric positions and Gabor wavelet coefficients) are extracted.  These features are fed in the input units of the two-layer perceptron.  The first layer aims to perform a nonlinear reduction of the dimensionality of feature space.  The second layer makes a statistical decision based on the reduced set of features.  An output unit is associated with a particular facial expression. Sari Meriem Structural Features 1ére Année Master GL
  • 28. Experiments 28  We use the cross-validation technique to test different configurations of our FER architecture.  We have carried out experiments on the FER using the developed architecture by using geometric positions alone, using Gabor wavelet coefficients alone, and by using the combination of the two pieces of information. Sari Meriem Structural Features 1ére Année Master GL
  • 29. Experiments 29  The recognition rate achieved by our system is 90.1% with 7 hidden units. Sari Meriem Structural Features 1ére Année Master GL
  • 30. Experiments After Excluding Fear Images 30  The system achieves a generalized recognition rate of:  73.3% with geometric positions alone.  92.2% with Gabor wavelet coefficients alone.  92.3% with combined information. Sari Meriem Structural Features 1ére Année Master GL
  • 31. Conclusion 31 In this paper we used of types of features in facial recognition:  Geometric features which are a set of positions of face components.  Recognition rate of 73.3% when used alone.  A multi-scale/oriented features called Gobor wavelet coefficients.  Recognition rate of 92.2% when used alone.  A recognition rate of 92.3% when combined together. Sari Meriem Structural Features 1ére Année Master GL