Accurate Hand Gesture Recognition using CNN and RNN Approaches

Abstract

Interaction between humans and computers, gesture recognit ion does play a critical role. By using modified convolutional neural network (CNN) and modified Recurrent Neural Networks (RNN) models for hand posture estimation, hand motion capture and hand object interaction in hand gesture recognition. By these models, we performed hand pose estimation on ICVL data set, motion capture available postures dataset and hand object interaction on kaggle dataset. RNN is for estimation and CNN is for feature extraction. A comparison of those three modules hand motion capture gives better accuracy.

Topics & Concepts