Publication details

Home Publications Publication details

Comparative Study and Evaluation of Six Face Recognition Algorithms with a View of their Application on Mobile Phones
Mahmoud N, Clarke NL
Advances in Communications, Computing, Networks and Security 6, ISBN: 978-1-84102-258-1, pp212-225, 2009
Can be ordered on-line.
Download links:  Download PDF

Mobile phones are rapidly becoming one of the most popular and powerful tools in our lives everywhere in the world (Clarke et al, 2003). They have provided a powerful ability whilst on move and increasingly sophisticated functions. Nowadays mobile phones are allowing people to access an increased amount of data and much more such as paying for products using micro-payments, surfing the Internet, buying and selling stocks, transferring money and managing bank accounts (Dagon et al, 2004). . However, security levels provided on the mobile phones these days such as PIN numbers and passwords do not provide substantial protection. This has highlighted the need for another strong way to protect the information being held on these devices as well as the services being served. It was proposed to implement methods of controlling access to these devices namely, biometrics. Face recognition is one of biometrics techniques can be implemented on mobile phones these days as most have integrated digital cameras which can be used to capture an image and used for authenticating legitimate users. It is based on the use of underlying algorithms to implement a solution. Six such algorithms cover provide a good coverage of the techniques available today were evaluated based on two experiments, a control experiment to evaluate the normal operating performance of the algorithms and a test experiment to test the ability of the algorithms to deal with facial images with varying facial orientation. The best performed algorithm in the control experiment was Gabor filters for face recognition with 4.5% misclassification rate and the best performed algorithm in the test experiment was Fisherfaces for face recognition with 35.1% misclassification rate.

Mahmoud N, Clarke NL