Robust hand gesture recognition based on mWLD and LTeP descriptors

Hand gesture recognition has been widely used to develop human-computer interfaces that are both efficient and intuitive. In this context, we propose in this paper a hand gesture recognition system based on two descriptors, namely, Local Ternary Pattern (LTeP) and multi-scale Weber Local Descriptor (mWLD), and for classification task, we used Support Vector Machine (SVM) and Radial Basis Neural Network (RBNN) classifiers. Two user independent datasets were used to evaluate the performance of our system, the first one is Jochen Triesch’s static hand posture dataset and the second is Sebastien Marcel’s dynamic gesture dataset. Obtained results show that LTeP descriptor performs better for static gestures achieving 91.67% recognition rate for Triesch’s dataset, whereas mWLD performed better for Marcel’s dataset where a 100% recognition rate was achieved.