Biometric identification system based on Principal Component Analysis

Abstract

The pattern recognition system for biometric identification, which was presented in this paper, used mathematical and statistical approaches such as Principal Component Analysis as a feature extraction method also Cross Validation and k-nearest neighbor with Euclidean metric distance for the classification method. The proposed recognition system used face and androgenic hair as biometric traits with total 400 images for both databases. Total images from each database were taken from 25 respondents and 16 images from each respondent. The highest precision achieved when the system used histogram equalization with 2-fold cross validation, 76.68 % of average precision for face database and 75.19 % of average precision for androgenic hair database. Both average of precision was obtained by using 90 most significant eigenvalues and its corresponding eigenvectors in the Principal Component Analysis feature extraction method.