Skeleton Based Dynamic Hand Gesture Recognition using Short Term Sampling Neural Networks (STSNN)

Abstract

This research introduces an innovative framework for real-time dynamic hand gesture recognition in the field of Human-Computer Interaction (HCI). The framework combines depth learning networks with the integration of multiple datasets to extract both short-term and long-term features from video input. A significant contribution of this research lies in the integration of Convolutional Neural Networks (CNNs) into a specialized short-term memory network (STSNN), enabling the capture of long-term contextual information for accurate gesture recognition. The proposed framework is thoroughly evaluated using two hand-held databases, namely the 14/28 dataset and the LDMI database. By leveraging the computational power of depth learning networks and the fusion of diverse datasets, our model outperforms previous methods, establishing its efficacy in real-time dynamic hand gesture recognition tasks. The outcomes of this research significantly contribute to the advancement of HCI, providing a robust and technically sophisticated solution for gesture-based interfaces. The findings hold promise for enhancing user experiences and facilitating seamless integration of gesture-based interaction techniques across various domains, ultimately improving the efficiency and effectiveness of human-computer interactions.