Two-class Discrimination of Personal Attributes on Gabor-filtered Features of Facial Image(International Workshop on Advanced Image Technology 2007)

概要

論文の詳細を見る
The human face provides important impressions of various biological attributes such as age and gender, and automatic recognition of such personal attributes by facial image processing has been attracting an increasing attention. In this study, we focused on gender as one example of an attribute, and investigate the advantages of the latest classifier of note, the Support Vector Machine (SVM), over one of the classic classifiers, Fisher's linear discriminant, in terms of performance in a two-class gender discrimination task. Experimental results showed that the accuracy of gender classification achieved over a test set of 100 facial images reached 95% by using the SVM classifier, and the error rate was less than one-fourth that of the Fisher's linear discriminant classifier.
社団法人電子情報通信学会の論文
2007-01-03