Multimodal Classification using Feature Level Fusion and SVM

The use of biometrics in the field of enhancing security and authentication in sensitive systems is a rapidly evolving technology. The increasing attacks and decreasing security in unimodal systems have resulted in designing multimodal systems combining different biometric traits. A lot of research has already been done in designing multimodal systems with fusion at rank and match-score level using different classifiers such as Bayesian classifiers, LDA, ANNs and SVMs. In this research work, a multimodal system is designed by integrating face, fingerprint and palmprint based on feature level fusion. Each of the feature vectors are extracted independently using PCA and then fused together to perform classification using a multiclass SVM. The classification is performed on a set of test images taken from both the standard databases and live images captured in biometric lab.