An improved Genetic Optimized Neural Network for Multimodal Biometrics

Abstract

In this paper, a novel classification technique for multimodal biometric system based on fingerprint and palmprint is proposed. The problems faced in unimodal biometric system such as noisy data, intra class variations, restricted degrees of freedom, non-universality, spoof attacks, and unacceptable error rates are overcome in multimodal biometric system by integrating the evidence presented by multiple traits. It is proposed to fuse the features of the fingerprint with palmprint images. Features are

extracted using Gabor filter and Discrete Cosine Transform (DCT). The extracted feature vectors were classified using an improved Partial Recurrent Neural Network with genetic optimization. The proposed Momentum Optimized Genetic Partial

<span style="font-size:12.0pt;font-family:" times="" new="" roman";mso-fareast-font-family:="" "times="" roman";mso-ansi-language:en-in;mso-fareast-language:en-in;="" mso-bidi-language:ar-sa"="">Recurrent Neural Network (MOG-PRNN) was evaluated using a publicly available dataset and features obtained from live dataset. The experimental results obtained show an average classification accuracy of 98.6% with different datasets.</span>