Improving the Performance of Facial Recognition System Using Artificial Neual Network
- June 23, 2021
- Posted by: rsispostadmin
- Categories: Computer Science and Engineering, IJRIAS
International Journal of Research and Innovation in Applied Science (IJRIAS) | Volume VI, Issue V, May 2021|ISSN 2454-6194
Improving the Performance of Facial Recognition System Using Artificial Neural Network
Asogwa Tochukwu Chijindu, Ugwu Edith Angela, Mbah David Luchi
Computer Science, Enugu State University of Science and Technology, Enugu, Nigeria
Abstract
This work presents “improving the performance of facial recognition system using artificial neural network”. The aim is to develop a more reliable and précised face recognition system. This will be achieved using the AT&T database as the training dataset, image acquisition, image processing, and artificial neural network. The work will be implemented using image processing toolbox, image acquisition toolbox, statistics and machine learning toolbox and Mathlab. The accuracy was measured using the neural network performance evaluation toolbox and the result achieved is 97.6%.
Keywords: facial recognition, artificial neural network, accuracy, AT & T dataset
I.INTRODUCTION
According to [1], facial recognition is dated back in the early 70’s. In simple words, it is a process of identifying and recognizing a face. But as easy as it sounds is not a simple task because there are lots of similarity in certain facial traits of different individuals, and this similarity in face has resulted to lots of problems like mistake identity, mistake arrest, fraud, impersonation and lots more. These challenges have become a major problem in the world generally.
Today face recognition has drawn the attention of researchers in fields from security, identification, engineering, banking sector, psychology, and image processing, computer vision and lots more. In fact a search of facial recognition on google.com produces over 350,000 000 results. However, one will wonder why despite the numerous works done and huge success achieved so far with this technology, yet research works are still being published on the topic daily. The answer to this is simply lack of consistency in the performance accuracy of the conventional systems.
