Multimodal Biometric Identification System using Deep Learning

Abstract

In real world we knows that a multimodal biometric system performs better and overcomes the limitation and gives better classification accuracy when compare to than Unimodal biometric system. This paper proposes a novel multimodal biometric recognition system based on feature-level fusion and deep learning model. The significance of this paper is, it focuses on the issue of selection of best feature extraction and classification techniques, by investigating different types of feature extraction techniques with different databases of given modality like face, Plamprint and iris. We proposed unimodal biometric recognition using Convolution Neural Network (CNN). Later the results of unimodal recognition used two-layer fusion to build multimodal biometric recognition. Features like Historgram of Gradient, Zernike Moments and Pseudo Zernike Moments are extracted. The performance of proposed multimodal recognition method shows better recognition accuracy than unimodal recognition.