Hand Gesture Recognition System using Modified SVM and Hybrid Ensemble Classifier

Hand Gesture Recognition (HGR) methods have gained tremendous interest in the past few years. The technique of HGR allows humans to connect with the system and interact naturally, thereby avoiding the involvement of any mechanical amenities. Automatic control of home appliances in smart home is an important application of a HGR system. In this paper, we propose a new HGR system using Speeded Up Robust Feature (SURF) as the feature descriptor. Bag of Feature (BoF) algorithm is employed to generate visual histogram of the SURF features and to generate a unified vector by mapping to the visual vocabulary. The initial stage of classification is performed by the proposed modified Support Vector Machine (SVM) classifier. In the second stage a classifier fusion model called as hybrid ensemble classifier obtained by combining K-Nearest Neighbour (KNN) and the modified SVM classifier is used. The experimental results show that the proposed hybrid ensemble and the modified SVM classifier provides better results compared to the individual classifiers.