Ensemble Learning for Static Hand Gesture Recognition using HOG and LBP Features on RGB-D Data

Abstract

Hand Gesture Recognition (HGR) systems has gained a lot of interest in research community due to its application in Human Computer Interaction (HCI), Advanced Driver Assistance Systems (ADAS) and Sign Language Recognition (SLR) for non verbal communication using various hand postures. Multi modal HGR systems with combination of RGB, depth and sensor data etc., have proved to be more efficient as compared to uni-modal systems. Also the classification decision based on the voting of different classifiers can be more accurate than single classifier. In this paper, we propose ensemble classifier of support vector machine, random forest and multi-layer perceptron classifiers for classification of hand gestures. Ensemble classifier is evaluated on HOG, LBP features with principal component analysis (PCA) and the pre-trained VGG16 model based deep features on both RGB and Depth data. Experiments are conducted on two different RGB-D dataset NTU and OUHANDS to evaluate the proposed method. Average classification accuracy of 97.50% is achieved on NTU dataset using the proposed method.