2D-3D FACIAL EXPRESSION RECOGNITION BU GRAPH FORMING /
Akhgar, Mahsa
2D-3D FACIAL EXPRESSION RECOGNITION BU GRAPH FORMING / MAHSA AKHGAR; SUPERVISOR: Asst. Prof. Dr. PAYAM ZARBAKHSH - xi, 84 sheets: photos; 31 cm. Includes CD
Thesis (MSc) - Cyprus International Univesity. Institute of Graduate Studies and Research Electrical and Electronics Engineering
Includes bibliography (sheets 77-84)
ABSTRACT
The need for biometric data analysis is an integral part of today’s digital world. With
advances in data registration and machine learning techniques, developing image
processing technologies such as identity recognition and facial expression detection
has attracted a significant intersect among the research and tech community. Emotion
detection from facial images has many applications ranging from surveillance and
security to psychology and marketing. These systems, however, are prone to errors due
to variable factors such as brightness intensity or changes in facial tissue. In this study,
a novel graph forming method based on invariable points of human face is proposed.
These points are called invariable because they are robust to the changes in brightness
and texture. Emotions are manifested in the proposed graph-based features extracted
from 2D and 3D images. On top of the feature extraction component, we also assessed
the effect of applying a feature selection method based on minimum redundancy and
maximum relevance. Three well-known classifiers namely SVM, Random Forest (RF)
and kNN are used to detect the labels. The results confirm the effectiveness of the
feature selection stage. Moreover, the findings imply that graph-based features from
3D images result in a better performance. When comparing the classifiers, SVM is the
winner followed by the RF. The best results are achieved when graph-based features
from 2D and 3D images are concatenated. In overall, the recognition rate of the
proposed method is superior to the state of the art.
Keywords: 2D+3D Images; Facial Expression Recognition Geometric; Facial
Features; kNN Classifier; RF Method; SVM classifier.
Facial expression--Dissertations, Academic
Three-dimensional imaging--Dissertations, Academic
2D-3D FACIAL EXPRESSION RECOGNITION BU GRAPH FORMING / MAHSA AKHGAR; SUPERVISOR: Asst. Prof. Dr. PAYAM ZARBAKHSH - xi, 84 sheets: photos; 31 cm. Includes CD
Thesis (MSc) - Cyprus International Univesity. Institute of Graduate Studies and Research Electrical and Electronics Engineering
Includes bibliography (sheets 77-84)
ABSTRACT
The need for biometric data analysis is an integral part of today’s digital world. With
advances in data registration and machine learning techniques, developing image
processing technologies such as identity recognition and facial expression detection
has attracted a significant intersect among the research and tech community. Emotion
detection from facial images has many applications ranging from surveillance and
security to psychology and marketing. These systems, however, are prone to errors due
to variable factors such as brightness intensity or changes in facial tissue. In this study,
a novel graph forming method based on invariable points of human face is proposed.
These points are called invariable because they are robust to the changes in brightness
and texture. Emotions are manifested in the proposed graph-based features extracted
from 2D and 3D images. On top of the feature extraction component, we also assessed
the effect of applying a feature selection method based on minimum redundancy and
maximum relevance. Three well-known classifiers namely SVM, Random Forest (RF)
and kNN are used to detect the labels. The results confirm the effectiveness of the
feature selection stage. Moreover, the findings imply that graph-based features from
3D images result in a better performance. When comparing the classifiers, SVM is the
winner followed by the RF. The best results are achieved when graph-based features
from 2D and 3D images are concatenated. In overall, the recognition rate of the
proposed method is superior to the state of the art.
Keywords: 2D+3D Images; Facial Expression Recognition Geometric; Facial
Features; kNN Classifier; RF Method; SVM classifier.
Facial expression--Dissertations, Academic
Three-dimensional imaging--Dissertations, Academic