Hand pose estimation based on deep learning depth map for hand gesture recognition

Hand pose estimation plays an important role in many applications, especially in human-computer interaction. Therefore, this topic has matured quickly in recent years. In this work we focus on the hand pose estimation from a depth map using convolutional neural networks. We propose a method for hand pose estimation by formulating a regression problem whose solution is the 16 hand joint locations. This method consists of two stages, the first one dealing a hand detection based on contours, the second one consists hand pose estimation using con-volutional neural networks. In this paper, we provide an extensive quantitative and qualitative experiments using real word depth maps from ICVL dataset. We perform a comparative evaluation with the state-of-the-art approaches to show the effectiveness and the accuracy of our method. Moreover, we propose a new application for hand gesture recognition based on our hand pose estimation method. The experimental results reported on test sequences of ICVL dataset show that the proposed application yields interesting performances and gives a marked improvement in recognition rate.