Multispectral fingerprint biometrics

Abstract

A novel fingerprint sensor is described that combines a multispectral imager (MSI) with a conventional optical fingerprint sensor. The goal of this combination is a fingerprint sensor with improved usability and security relative to standard technology. The conventional sensor that was used in this research is a commercially available system based on total internal reflectance (TIR). It was modified to accommodate an MSI sensor in such a way that both MSI and TIR images are able to be collected when a user places his/her finger on the sensor platen. The MSI data were preprocessed to enhance fingerprint features. Both the preprocessed MSI images and the TIR images were then passed to a commercial fingerprint software package for minutiae detection and matching. A multiperson study was conducted to test the relative performance characteristics of the two types of finger data under typical office conditions. Results demonstrated that the TIR sensor performance was degraded by a large number of poor quality fingerprint images, likely due to a large percentage of samples taken on people with notably dry skin. The corresponding MSI data showed no such degradation and produced significantly better results. A selective combination of both modalities is shown to offer the potential of further performance improvements.