Hand gesture recognition in ASD related motion using YOLOv8 algorithm / Muhammad Afiq Mohd Ali

The study focuses on designing a Hand Gesture Recognition application by means of the YOLOv8 deep learning algorithm to recognize hand gestures associated with autism spectrum disorder (ASD). The study aims to find out how effective YOLOv8 would be for such hand motions and to train the model with a custom dataset of hand gestures. The prototype developed for evaluation tested the system with Precision, Recall, and Mean Average Precision (mAP)-the F1 score was 83%, certifying further the operational capacity of gesture recognition associated with ASD. The results demonstrate how this can assist in making proper diagnosis and skyrocket systematic understanding into unexplored domains of ASD. Inadequate availability of an exhaustive dataset for ASD motions and the unsuitability of the system to understand intricate ASD-related gestures were flagging hurdles. Engaging into a heterogeneous dialogue with autistic specialists would boost the position of computer scientists on this issue and yield contact to the improvement of more accurate and diverse datasets. To enhance recognition accuracy, integration of other motion features that involved body posture and facial expressions should be completed. Future work should help develop protocols that expand dataset availability and increase the model's accuracy while incorporating some complex AI algorithms to better support these fields in providing clinical and therapeutic applications.