Hyperspectral biometrics for facial mode: An alternate approach to multimode method

Abstract

Use of Biometric systems has been increased in past decades due to increasing demands of security. The unimodal biometric system has to suffer from problems such as intra class variation, noise in the sensor data etc. This problem can be solved using multimodal biometric fusion. In this paper authors have compared the Hyperspectral and multimodal (fingerprint and face) fusion at matching score level. The Hyperspectral images were fused for combinations like 650+710nm, 650+710+770 nm and the performance parameter like False Non Match Rate (FNMR) and False Match Rate (FMR) have been performed. The performance parameters for Hyperspectral imagery outperform that of the facial mode for a single spectral band i.e. 650nm. Also parameters for the Hyperspectral fusion of 650+710nm, 650+710+770nm combinations spectral modes are promising and are comparable for higher order combination to that of multimodal fusion biometrics. It is observed that the Hyperspectral fusion for higher spectral bands combinations is linear improvements in Equal Error Rate (EER) percentage.