AUTOMATED GESTURE RECOGNITION USING ARTIFICIAL RABBITS OPTIMIZATION WITH DEEP LEARNING FOR ASSISTING VISUALLY CHALLENGED PEOPLE

Gesture recognition technology has become a transformative solution to enhance accessibility for people with vision impairments. This innovation enables the interpretation of body and hand movements, transforming them into meaningful information or commands by applying advanced computer vision sensors and algorithms. This technology serves as an intuitive interface for the visually impaired, enabling them to access information seamlessly, navigate digital devices, and interact with their surroundings, fostering more independence and inclusivity in day-to-day activities. Gesture recognition solutions using deep learning (DL) leverage neural networks (NN) to understand intricate patterns in human gestures. DL algorithms can identify and classify different hand and body movements accurately by training on extensive datasets. Therefore, this study develops an automated gesture recognition using artificial rabbits optimization with deep learning (AGR-ARODL) technique for assisting visually challenged persons. The AGR-ARODL technique mainly intends to assist visually challenged people in the recognition of various kinds of hand gestures. In accomplishing this, the AGR-ARODL technique primarily pre-processes the input images using a median filtering (MF) approach. Next, the AGR-ARODL technique involves the SE-ResNet-50 model to derive feature patterns and its hyperparameter selection process is carried out by the use of the artificial rabbit optimization (ARO) algorithm. The AGR-ARODL technique applies a deep belief network (DBN) model for the detection of various hand gestures. The simulation results of the AGR-ARODL method are tested under the benchmark gesture recognition dataset. Widespread experimental analysis underscored the betterment of the AGR-ARODL technique compared to recent DL models.