Linearly Separable SubCode: A novel output label with high separability for biometric discretization

Abstract

Separability in a code is crucial for guaranteeing a decent separation between multiple classes in terms of Hamming distance. In biometric discretization, this property is of particular useful for preserving dissimilarity in the continuous feature space when a feature is mapped to a code space. Binary Reflected Gray Code (BRGC), a widely employed output label in discretization schemes, however, does not appear to exhibit such a property, yielding an indefinite mapping process that impedes the overall recognition performance. This paper explores and analyzes the separability of BRGC and put forward a class of coding scheme which exhibits high separability capacity.