Shape-based hand recognition approach using the morphological pattern spectrum

bstract. We propose the use of the morphological pattern specrum, or pecstrum, as the base of a biometric shape-based hand ecognition system. The system receives an image of the right hand f a subject in an unconstrained pose, which is captured with a ommercial flatbed scanner. According to pecstrum property of inariance to translation and rotation, the system does not require the se of pegs for a fixed hand position, which simplifies the image cquisition process. This novel feature-extraction method is tested sing a Euclidean distance classifier for identification and verificaion cases, obtaining 97% correct identification, and an equal error ate (EER) of 0.0285 (2.85%) for the verification mode. The obtained esults indicate that the pattern spectrum represents a good featurextraction alternative for lowand medium-level hand-shape-based iometric applications. 2009 SPIE and IS&T. DOI: 10.1117/1.3099712