"Human Hand Gesture Recognition Based Feature Extraction and Classification Using Sign Language Data"

Hand gestures are natural forms of Human-Computer Interaction (HCI).Hand gesture recognitions are rapidly developing fields of study in computer vision and machine learning as they offer wide ranges of potential applications, allowing people to interact with robots and system interfaces in more straightforward and natural ways without requirements for additional hardware.Hence, gesture recognition studies aim to develop systems that can recognize certain human gestures and utilize them to operate devices or transmit information.In order to accompolish real-time gesture recognitions, reliable hand detections are necessary for vision-based hand gesture interfaces.The initial part of this work is data collection, and standard dataset is introduced for data validations.Dataset with 26 signs was created using Sign Language (SL) and assessed in real time for a variety of lighting scenarios.The suggested technique recognises motions and identifies human hands using a feature set of symmetric patterns that is obtained using techniques for extracting hand perimeter features.Depth based Speeded-Up Robust Features (DSURF) is employed to identify important focal spots in SL-based images along with their associated characteristics.Finally, hand gestures are classified using Improved Graph Neural Networks (IGNN) trained on symmetric patterns under varying illumination conditions.Selected characteristics are validated where experimental results suggested that this work's schema outperformed other methods in twerms of precision for a variety of illumination conditions and ranging pixel sizes.