Hand pose identification from monocular image for sign language recognition

In this paper, a novel approach for hand pose recognition is proposed by analyzing the textures and key geometrical features of the hand. A skeletal hand model is constructed to analyze the abduction/adduction movements of the fingers and subsequently, texture analysis is performed to consider some inflexive finger movements. Probabilistic distributions of the geometric features are considered for modelling intra-class abduction/adduction variations. Gestures differing in inflexive positions of fingers are classified based on Homogeneous Texture Descriptors (HTD), where the texture region is characterized using the mean energy and energy deviation from a set of frequency channels. Similarity measures are computed between input gestures and pre-modelled gesture patterns from a database by considering intra class abduction/adduction angle variations and inter class inflexive variations. Experimental results show the efficacy of our proposed hand pose recognition system.