Non-negative sparse coding based scalable access control using fingertip ECG

This work evaluates feasibility of using electrocardiogram (ECG) signals from fingertips for biometrics. These non-intrusive easily acquired signals are tested for performance in verification mode using a large publicly available database containing fingertip ECG. Two methodologies are presented for the design of an authentication system for scalable access control wherein we propose use of Non-Negative Sparse Coding followed by Linear Discriminant Analysis. Furthermore, based on the application scenario, two classifier methods are proposed. We compare our system with AC/LDA for performance in large-scale deployment scenarios and scalability. For our methods, promisingly low equal error rates are obtained - in particular, 2.59% for a population size of 1012 users.