A Soft Computing Model of Soft Biometric Traits for Gender and Ethnicity Classification

Abstract

There is paucity of information on the possibility of ethnicity identification through fingerprint biometric characteristics and so, this work is set to combine two soft biometric traits (Gender and Ethnicity) in order to ascertain if individual of different ethnicity and gender bias can be identified through their fingerprint. Live scan mechanism was used in order to minimize human errors and as well speed up the rate of fingerprint acquisition which unequivocally ensure good quality capturing of the fingerprint image. In this work, fingerprints of over a thousand people from three different ethnic groups of both male and female gender in Nigeria were captured and subjected to training, testing and classification using Gabor filter and K-NN respectively. Histogram equalization was used for image enhancement and the system performance was evaluated on the basis of some selected metrics such as Recognition Accuracy, Average Recognition Time, Specificity and Sensitivity. Result of this work indicated over 96% accuracy in predicting person’s ethnicity and gender with an average recognition time of less than 2secs. Index