Eigennose: Assessing Nose-Based Principal Component Analysis for Achieving Access Control with Occluded Faces

Abstract

State-of-the-art face recognition systems exist today with varying performances. However, many suffer from multiple occlusions that threaten their performance. The common causes of these occlusions are hats, scarves and, sunglasses. Usually, when occlusions are present, the nose features are available. Surprisingly, not much research has been focused on nose biometrics. Research has shown that the nasal area provides robust, discriminant features that can be used to positively authenticate a user. In our system, we attempt to authenticate a user using only their nose. Eigennose algorithm, which is an extension of the eigenface algorithm is developed to find the discriminant nasal features of individuals with Euclidean distance used for matching. The system is then compared with machine learning algorithms such as Support Vector Machines and k-Nearest Neighbor to find better-performing methods. Our experiment did not achieve very good performance.