Multimodal Biometric Invariant Fusion Techniques

Abstract

The hand geometry, features in face, iris scan, and fingerprint vary from person to person, which provide unique features to be used in biometrics field for providing security to various systems. Most of the mono-biometric authentication systems give high error rate as they use only one feature. Hence, multimodal biometric systems are introduced, which can help in reducing the error rate at the cost of maintaining more data related to the features. Hence, it is said to be that the multimodal biometric systems are more reliable and secure. Image-based approaches offer much higher computation efficiency with minimum preprocessing. This approach is proved to be effective as the reliable feature extraction is possible even when the quality of image is low. However, this approach is weak if there are distortions in the shape of the image and variation in the positions or the orientation angle. Hence, this chapter presents a multimodal biometric invariant fusion authentication system based on fusion of Zφ invariant moment of fingerprint and face features. It reduces the storage of more features for authentication and reduces the error rate. The Morlet wavelet transform is used to make the system less sensitive to shape distortion by smoothening and preserving the local edges. The Zφ moment is the combination of Zernike and invariant moments, which are used to produce an affine transformation that is extracted from the fingerprint and the face. Authentication is successful if the similarity is 90% in the case of fingerprint and 70% in the case of face. False acceptance rate (FAR) and false reject rate (FRR) are optimal with these threshold values.