Research
Thesis
-
Cognitive Vision-Based Hand Gesture Recognition System for Media and Accessibility Control
Abstract
This project discusses a real-time cognitive vision system for gesture based media control by leveraging webcam based hand and finger tracking. The goal of the system is enhancing access, and interaction for those with physical disabilities by providing touch-free control to media actions like: volume adjustments, screen brightness control, play and pause. By utilizing computer vision and low-power machine learning models, the system decodes finger landmarks from a conventional webcam in real time for high accuracy continuous gesture recognition with low latency. Continuous scale-volumes are associated with (e.g.) the distance between thumb and index-finger, conveying properties like size or amount of an item, and finite control-expression templates corresponding to postures on the hand. The system was planned to be able to perform steadily and in realtime at rates of 30-60 FPS, and remain robust with performance reduction as a result of different users and the variation in lighting conditions. This study contributes to the area of cognitive HCI by offering an inexpensive interactive camera-based assistive technique, which goes beyond normal conventional input interactions. This project will display the combination of computer vision and cognitive modeling.