期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2016
卷号：92
期号：1
出版社：Journal of Theoretical and Applied
摘要：In Biometrics, physiological or behavioral features are utilized to validate an individual's identity. Though, a substantial amount of research has carried out in this field, unimodal biometric frameworks regularly experience some limits because of non-universal biometrics attributes, vulnerability to biometric spoofing or lack of accurateness. In this paper, the accuracy problem is addressed through multimodal biometric fusion. Our proposed multimodal biometric fusion methodology offers face and fingerprint as biometric traits as an input for sanctuary purpose that are not unique to each other of the human body. Here, we include Wiener filter for preprocessing phase and Discrete Wavelet Transform (DWT) for the fusion process of the two traits. Also, a linear discriminant regression classification (LDRC) algorithm has been proposed. We propose selective small reconstruction error (SSRE) which helps to select the classes, wherein chances to misclassifies the classes are considered when calculating the between-class reconstruction error (BCRE). After maximizing the ratio of BCRE and within-class reconstruction error (WCRE) an optimum projection matrix is obtained through which a high discrimination value can be achieved for classification. Finally, the experimentations are carried out and our proposed LDRC methodology performance is found better than the existing LRC in terms of accuracy, FAR, FER, EER.
