Real Time Detection on Face Side Image with Ear Biometric Imaging Using Integral Image and Haar-Like Feature

Abstract

There are no CCTV that accurately record human movement from the front of camera. Sometimes the recording shows the figure of a human face from the side. Face detection is usually used to identify facial data with different expressions. Various studies have been developed on Facial Recognition and Expression. However, most of the analysis is about the face from the front side. To deal with the problem, it is necessary to develop research on the side face using ear biometric. This research uses Integral Image, Adaboost and Haar-Like techniques. Direct test via laptop webcam with the object facing from the side face. The experiment result of this research is that the system successfully classifies 33 ears that can be detected from 60 ear database which has been labeled manually. From 33 ear detected there are 29 ears that can be classified correctly. So the accuracy is 88% as false positive. Fault tolerance of ear position is 45.07/345384 or 0.01% in data testing. The use of Haar-Like Feature has been successfully used as a biometric ear detector on the side of the face images. Ear detection takes place at 60 frames in 5 seconds. Or about 12 frames per second. For testing via webcam, the ear can be detected in about 0.5776376 seconds.