Hand Gesture Recognition System Based on Textural Features

Abstract

Gesture is an effective mean of communication between deaf or hearing-impaired people. Designing a hand gesture recognition system involves an efficient characterization step in order to reduce the amount of information contained in the images. We propose two descriptors known as Binary Pattern of Phase Congruency (BPPC) and Monogenic Binary Coding (MBC) to characterize our hand images. Three datasets are used for this purpose, the American, Arabic and dynamic dataset. Support vector Machine SVM and Radial Basis function Neural Network RBF NN are used to build our hand gesture recognition system. 94 to 96% of accuracy was obtained for the three datasets.