Facial expression, visual interactions, through which humans shows their inner emotional state,
thus it plays an important role in social interaction and interpersonal
relations. Facial expression recognition plays a very huge role in terms of
human to computer interaction as well as various aspects of behavioral science.
There are six known classes of emotional state (Angry, Disgust, Fear, Happy,
Sad, Surprise) associated with their respective facial expressions, according
to Ekman’s studies. Humans recognize facial expressions almost effortlessly and
without delay, but this is quite challenging for machines. This thesis presents
Facial Expression Recognition Using Enhanced Local Binary Patterns which uses
LBP for feature extraction.  The main
contribution of the thesis is the enhanced LBPs, in which the high variance LBP
pixels are selected to represent facial information and its recognition rates
outstandingly improved.  The tests was completed
on the BU-3DFE database. The experiments show that after applying feature
selection to enhanced LBP representations, the recognition rates are improved
by 11.67%. 

Keyword:
expression recognition, BU-3DFE, feature extraction, enhanced local binary
pattern.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

 

1.   
INTRODUCTION

Facial
Expression Recognition FER, a very import aspect of computer vision, is some
biometrics that seeks to use computational algorithms to detect expressions of
faces from an existing set of images in dataset. It has become one of the most
popular biometric and challenging topic in pattern recognition as much progress
has been made in respect to 2D images 12 3. Facial expression on the other
hand is a visible exhibition of emotion, intention, cognitive activity and
psychopathology of an individual 4; by means of adjusting the facial
muscles for each level of facial expression. Humans can easily understand each
facial expression, whereas it is difficult for machines to recognize faces or
face expression. Advancement and research in algorithms have developed methods
for identifying faces from given image and most fortunately, identifying,
classifying and recognizing emotional expressions from digital image.

 

There are three main steps in
facial expression recognition 29: facial acquisition, facial feature
extraction, and facial expression classification. Facial acquisition is a
preprocessing stage in which the input images region is been detected or
located, once a facial image is been located, 
the eye distance and the gray level of the corresponding face regions
will be normalize to be the same while the facial feature extraction will
concentrate more in finding the appropriate representation of the facial images
for the recognition and two main approaches in facial feature extraction:
appearance features-based systems and geometric features-based systems.
Appearance Features-Based Systems: it checks the changes in appearance (facial
images) such as wrinkled face, furrows and bulges. Image filters such as Gabor
wavelet analysis 35,36 and local binary patterns (LBP) 34 will be applied
to either specific face regions or the whole face to extract the facial
appearances changes. Gabor wavelet analysis is actuary high and its
computational memory requirement is very large while local binary patterns
(LBP) is more tolerance against illustration changes which is one of its
important properties also its simplicity in computation, it also have extensive
interest for facial feature expression representation. Geometric features-based
systems: detects the shapes and locations of major facial components such as
mouth, nose, eyes, and brows of the image Moreover, in practice Geometric
features-based (GFB) system require the accurate and more reliable facial
feature detection, which in real-time application is difficult.

 

 

Robust
algorithm has to be applied on each stage (image acquisition or registration,
normalization, feature extraction, classification and recognition) of the Facial
Expression Recognition (FER) system and novel methods used to recognize each
emotion with varying level of intensity. 
Researches are making effort to extract different facial feature from
different expression level, but their recognition performance generally depend
upon the reliability of these feature 5. For this reason, this paper made
use of the popular algorithm, Local Binary Patterns (LBP) for feature extraction and classified using distance
classifier. Since the Local Binary Patterns (LBP) has recently gained attention
in the field of facial recognition, so it is worthwhile, using it in facial
expression recognition applications 3 6.

 

However,
the major aspect this project is most concerned about when dealing with Facial
Expression Recognition (FER) is the expression on the face 4. This is due of the challenging
fact that the human faces carry a lot more information such as the identity of
an individual, we
therefore need to find a way to remove the personal identity while working with
expression recognition.

 

Facial
Expression Recognition (FER) has application in areas such as human computer
interaction, computer vision and pattern recognition due to its application in
several areas such as customer satisfaction framework, in security system for
verification and authentication and some robots have been developed to benefit
from the ability to recognize facial expressions 7. Also,
the behavioral science or medicine are key areas that can take advantage of the
application of facial expression analysis 4.

This paper is organized as follows:
the most popular and successful Local Binary Patterns (LBP) operators in
section 4.1. Section 5 presents extraction of facial feature based on Local
Binary Patterns (LBP). Section 5.1 explains the face recognition scheme used.
Simulation result and conclusion are shown in sections 6 and 7 respectively.