Condução de cadeira de rodas usando rastreamento ocular por imagem

• Ano de defesa: 2020
• Autor(a) principal: Santana, Eder Manoel
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Engenharia Biomédica
• Palavras-chave em Português: Rastreamento ocular; Controle alternativo; Visão computacional; Tecnologia assistiva; Cadeira de rodas motorizada; Eye tracking; Alternative control; Computer vision; Assistive technology; Motorized wheelchair; CNPQ::ENGENHARIAS::ENGENHARIA BIOMEDICA; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::SISTEMAS DE COMPUTACAO; Engenharia biomédica; Cadeiras de rodas
• Link de acesso: https://repositorio.ufu.br/handle/123456789/30257; http://doi.org/10.14393/ufu.di.2020.651

The act of driving a motorized wheelchair becomes extremely difficult for people with severe motor disabilities. The ability to drive such a device independently requires specific motor, visual and cognitive skills, which in many cases have been significantly affected or even lost, making it impossible for the individual to have an independent life. In this perspective, the present work proposes a tool for driving a motorized wheelchair by means of eye tracking from the user's image registered by a notebook webcam. Through this strategy it was possible to dispense with the connection to the notebook of other peripherals traditionally used to perform the task, as is the case of eye tracking devices available on the market. For the development of the work, a search for related works was carried out in order to identify the existing technologies, as well as the main needs and difficulties of users with special needs. Then, the system was developed using the Gaze Tracking and Dlib libraries, which allow identifying the region to which the user's eye is directed. Which, in turn, was associated with the directional commands necessary for driving the chair. Finally, the system was tested by five wheelchair users and five healthy volunteers in order to evaluate it, both in terms of usability while driving and in terms of applicability to various pathologies. The results showed the viability of the proposed solution for the desired application. Thus, it is expected that in the future the developed system can be used by individuals to drive motorized wheelchairs effectively and safely.