A unified distance measure scheme for orientation coding in identification

For a recognition task, distance measure between patterns is one of the most important procedures. Orientation coding-based methods have achieved high accuracy and speed in recognition, such as competitive code, palmprint orientation code and robust line orientation code. For orientation coding-based patterns, OR_XOR (Hamming distance) and SUM_XOR (Angular distance) are usually used for measuring the distance. However, little work have been done to study the physical significance of these distances for orientation coding-based patterns. The relationship between SUM_XOR and OR_XOR is not clear. This paper is focusing on the relationship between them and proposes a unified distance measure (UDM) scheme. To automatically compute the parameters introduced in UDM model, the population based heuristic algorithms, particle swarm optimization (PSO), differential evolution (DE) and fireworks algorithm (FWA) are used for the determination of parameters. Experimental results on artificial date set, PolyU palmprint database and the collected finger-vein database, suggest that the proposed PSO-UDM, DE-UDM, FWA-UDM gain lower equal error rate.