Towards Human Action Recognition in Open Environments: A Proof of Concept on Spanish Sign Language Vowels
Vanessa Alvear, César Domínguez, Gadea Mata · 2024
Gestures and body language are fundamental communication mechanisms between human beings. Trying to make machines recognize this language is an active area of research in the field of computer vision that has applications in fields as diverse as video surveillance and healthcare. Usually, most of these developments are carried out in closed and controlled environments (e.g., a laboratory with specific conditions of light, climate, furniture, backgrounds, and so forth). Real-time motion and gesture recognition through cameras in an uncontrolled environment, open environments, or outdoor environments (such as a shopping mall, museum lobby, train station platform, or airport terminal) and allowing interaction with passers-by in these spaces is not so much studied. This task presents challenges related to the technological, social, and legal factors that need to be carefully addressed. This work is a first step towards human action recognition in open environments adapted to interaction without physical contact using videos captured with a camera in real-time in these spaces. In this manner, as a proof of concept we explore the development of a system that allows real-time recognition of Spanish Sign Language (Lengua de Signos Española - LSE), in particular, we select the vowels as gestures.