Hand gesture identification and recognition using modern deep learning algorithms

Abstract

In this paper, we proposed an approach towards the real-time hand gesture recognition using the Gaussian Mixture-based Background/Foreground Segmentation Algorithm. We proposed a method for feature extraction by using measurements on joints of the extracted skeletons. The proposed algorithm will build a background subtract model to get the foreground image. We applied Gaussian blur to the foreground image and threshold for binary images. The contour hull and convexity are used to build a 3D image of the hand gesture recognition. We constructed a dataset and defined the gestures. We trained them by gesture classifiers by some assumptions such that those can be easily understood. Experimental results proved the effectiveness and potential of our modern deep learning approach.