University Links: Home Page | Site Map
Covenant University Repository


Oladipo, Oluwasegun and Covenant University, Theses (2022) BLACK FACE-BASED AGE ESTIMATION SYSTEM USING GENETIC ALGORITHM-ARTIFICIAL NEURAL NETWORK. ["eprint_fieldopt_thesis_type_phd" not defined] thesis, Covenant University Ota.

[img] PDF
Download (635kB)


Age estimation is the determination of a person’s age based on its biometric features. Its application can be seen in areas such as forensic analysis. E-Government, security and surveillance. Face biometric being non-intrusive is a preferred biometric for age estimation. Research focus in face-based age estimation system is to increase the number of correctly classified images, reduce the recognition time, and make appropriate choice of feature extraction technique to use, especially in an uncontrolled environment. Some of the computational approaches that have been used for face-based age estimation include machine learning techniques such as support vector machines, neural network and Bio-inspired Features (BIF). However, optimization techniques can be integrated into the classification module of age estimation systems to improve the overall performance. The back-propagation algorithm is the most popularly used algorithm for training a multilayer Artificial Neural Network (ANN). It is an efficient technique applied to classification problems, but still suffers drawback with complex problem space, as it has the tendency to converge at a local minimal point. This study is aimed at developing a face-based Age Estimation System (AES) using Genetic Algorithm - Back Propagation Artificial Neural Network (GA-ANN) for improved age estimation. The combination is motivated by the fact that Genetic Algorithm (GA) has the potential to traverse the entire search space while remaining time efficient. Hence, offsetting the aforementioned problem. The study implemented two feature extraction techniques namely Local Binary Pattern (LBP) and Gabor Wavelet (GW) separately, to deduce which is most suitable for black faces. Principal Component Analysis (PCA) was further applied to the feature vector generated for the second level feature extraction in order to remove redundant features. The system was trained and tested with a newly developed database containing 855 black faces taken in an uncontrolled environment using a mobile app and 500 faces from the FG-NET database. The system was trained 80% and tested with 20% of the database. The developed systems LBP GA-ANN (LBGANN) and GW GA-ANN (GGANN) were implemented in MATLAB programming environment. The study showed that the developed GA-ANN based AES (LBGANN and GGANN) performed better than standard back propagation Artificial Neural Network (ANN) based systems (LBANN and GANN) in terms of Correct Classification Rate (CCR) and recognition time, as it showed a correct classification rate of 94.97% and 92.11% respectively as against the standard ANN-based system that had 89.69% and 88.72% respectively. The developed system incurred more training time as it iterates through several GA generations. The study also showed that LBP feature extraction technique is more suitable for faces as it better encodes face texture information and morphological changes during growth than Gabor wavelet.

Item Type: Thesis (["eprint_fieldopt_thesis_type_phd" not defined])
Uncontrolled Keywords: Age estimation, face-based, artificial neural network, genetic algorithm, black face
Subjects: Q Science > QA Mathematics
Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions: Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science
Depositing User: Mrs Patricia Nwokealisi
Date Deposited: 18 Jul 2022 11:02
Last Modified: 18 Jul 2022 11:02

Actions (login required)

View Item View Item