Multi-Unit Feature Level Fusion Approach Using PPCA

Iris recognition system is considered to be one of the most reliable, accurate and stable security systems providing identification or authentication of an individual. In this paper, a multi-unit feature level fusion approach for iris based biometric system is proposed with the aim of improving the recognition accuracy even for inaccurately segmented iris images. From the existing analysis it is clear that for an iris based biometric system to become reliable and accurate, more emphasis should be given to the preprocessing and segmentation stage. In this work Daugman’s Integro-Differential Operator is applied on eye images taken in unconstrained situation, and iris region of interest is extracted without eliminating noise factors like eyelid and eyelash occlusions, specular reflections, non-circularity of iris, etc. A novel feature selection technique, Probabilistic Principal Component Analysis (PPCA) is applied to provide good recognition rate by working on missing values of inaccurately segmented iris images. Mean method of multi-unit feature level fusion methodology is proposed to improve feature selection accuracy. The proposed methodology is found to provide better recognition rate of 83.3% even for inaccurately segmented iris images, when experimented on MultiMedia University (MMU) dataset.