Hybrid features of spatial domain and frequency domain for person identification through ear biometrics

Abstract

Biometrics proved to be very efficient, more natural and easier for users than traditional methods of human identification. This paper presents an accurate biometric system based on human ear. Many features have been extracted in the spatial domain such as area of the ear, ear edge points, and widths of ear in different places. Those features have been extracted in the enrollment stages and stored as templates. Using a comparing technique such as Euclidean distance for each feature or for the whole features gives different correct recognition rates, which reaches 88.2%. Using spatial domain features as well as frequency domain features such as FFT and DCT coefficients raises our recognition rate to 92% of success. 100% of correct recognition can be achieved by using the average values of five samples instead of three samples for each person.