Maya V. Karki & Dr. S. Sethu Selvi International Journal of Biometrics and Bioinformatics (IJBB), Volume (7) : Issue (1) : 2013 58 Multimodal Biometrics at Feature Level Fusion using Texture Features Maya V. Karki mayavkarki@msrit.edu Faculty, MSRIT, Dept. of E&C MSRIT Bangalore-54, INDIA Dr. S. Sethu Selvi selvi_selvan@yahoo.com Faculty, MSRIT, Dept. of E&C MSRIT Bangalore-54, INDIA Abstract In recent years, fusion of multiple biometric modalities for personal authentication has received considerable attention. This paper presents a feature level fusion algorithm based on texture features. The system combines fingerprint, face and off-line signature. Texture features are extracted from Curvelet transform. The Curvelet feature dimension is selected based on d-prime number. The increase in feature dimension is reduced by using template averaging, moment features and by Principal component analysis (PCA). The algorithm is tested on in-house multimodal database comprising of 3000 samples and Chimeric databases. Identification performance of the system is evaluated using SVM classifier. A maximum GAR of 97.15% is achieved with Curvelet-PCA features. Keywords: Multimodal Biometrics, Feature Level, Curvelet Transform, Template Averaging, PCA Features and SVM Classifier. 1. INTRODUCTION Personal authentication systems built upon only one of the biometric traits are not fulfilling the requirements of demanding applications in terms of universality, uniqueness, permanence, collectability, performance, acceptability and circumvention. This has motivated the current interest in multimodal biometrics [1] in which multiple biometric traits are simultaneously used in order to make an identification decision. Depending on the number of traits, sensors and feature sets used, a variety of scenarios are possible in a multimodal biometric system. They include single biometric with multiple sensors, multiple biometric traits, single biometric with multi- instances, single biometric with multiple representations and single biometric with multiple matchers. Among all these scenarios, system with multiple biometric traits is gaining importance and this method itself is known as multimodal biometric system. Based on the type of information available in a certain module, different levels of fusion are defined [2]. Levels of fusion are broadly classified into two categories: fusion before matching also called as pre-classification which includes sensor level and feature level. Fusion after matching also called as post classification which includes match score level and decision level. Amongst these, fusion at feature level is gaining much research interest. Most of the existing multimodal systems are based on either score level or decision level fusion [3]. Match score is a measure of the similarity between the input and template biometric feature vector. In match score level, scores are generated by multiple classifiers pertaining to different biometric traits and combined [4]. In order to map score of different classifiers into a single domain, where they possess a common meaning in terms of biometric performance, normalization technique is applied to the output of classifier before score fusion. Gupta [5] developed a multimodal system based on fingerprint, face, iris and signature with score level