Static hand gesture recognition using multi-dilated DenseNet-based deep learning architecture

Abstract

Hand Gesture Recognition (HGR), a Human-Robot Interaction (HRI) create user interfaces to enhance safety and control. Several approaches are introduced to recognize static hand gestures effectively, however the classification efficiency suffered due to various lightings like natural light, artificial light and dark rooms increases the difficulty of accurate HGR. To tackle these problems, a Multi-Dilated Convolution-based DenseNet (MDCDN) architecture, a combination of multi-dilated convolution and DenseNet is proposed which extracts the features automatically. Finally, the benefits of high-level deep learning techniques are leveraged for the gesture recognition from hand. Python is used for architecture evaluation. The outcome of proposed is estimated in terms of accuracy, recall, F-measure, precision, etc, using ASL, ISL, Massey and HSR real datasets. Each dataset contains huge amount of gesture classes and their pictures have an equal amount of uniform and complex backgrounds. The proposed results are promising and provide better performance than existing methods.