Asian Transactions on Computers
Volume: 01, Issue: 04, September 2011
Designing and Implementing System of Real Time Face Detection and Recognition: Based on RBF
Eri Prasetyo Wibowo, Bonaventura Pinandito, Bheta Agus Wardijono and Akeda Bagus
Digital Object Identifier: ATC-40101045
Face recognition is the one of many interesting researches concerned by biometric system since several years ago. This interest is caused by the easy process to obtain the research data (face images), although there are still many factors that can affect a failure rate of identification in pattern recognition of face. The research about face recognition develops continuously with the growing human needs in the security field, human computer interaction, and so on. Today, the object of research is not only the static image, but also begins to the real-time moving images.
The paper explores the implementation of Radial Basis Function (RBF) method in making real-time face recognition application. The input data for the RBF network is obtained from the face localization process using Haar-like feature method, egmentation, feature extraction, and normalization using image processing algorithm to get the fiducial points of face; representing the characteristics of each human face. The RBF values are then processed to calculate the difference level with each target; resulting of network learning process. The method that is used in learning process network is LMS (Least-Mean-Square) and to measure the difference level used Euclidean Distance method.
From the test results of 10 respondents with several conditions, RBF method has been able to do face recognition with the recognition success rate of 90% - 100% with a record in database, 81.6667% - 94.2262%, with 3 records in database, and 80% - 86.3095% with 5 records in database. The conditions of sampling and testing time affect the percentage of success rate.
Full Text: PDF (754 KB)
© Copyright Asian-Transactions 2013 - All rights reserved.
Use of this website signifies your agreement to the terms & conditions