STUDY OF THE NAVIGATION OF A WHEELCHAIR USING COMPUTATIONAL VISION CONCEPTS

• Main Authors: Marcos Batista Figueredo, Roberto Luiz Souza Monteiro
• Format: Article
• Language: Portuguese
• Published: Universidade Estadual do Sudoeste da Bahia 2017-03-01
• Series: Revista Saúde.com
• Subjects: heelchair; Assistive Technology; Cascade Classifiers; Active Shape Models
• Online Access: http://www.uesb.br/revista/rsc/ojs/index.php/rsc/article/view/446/396

Mobility is an important means of social interaction that, besides allowing the accomplishment of several daily tasks, establishes a connection of the patient with the social and work universe. For people who have the so called paraplegias or tetraplegias, the wheelchair is an important means of exercising their citizenship. Several researches seek to make navigation simple and efficient, but, in general, the presented solutions have a great amount of sensing, intrusiveness and high cost. We propose a computational model that allows the navigation of a wheelchair using facial expressions. Unlike the works studied, we suggest a model that is based on two facial expressions: the pose of the head and the closing of the eyes, and only an input sensor, a USB camera. The model converts facial expressions into commands for navigating the chair, and two techniques make interpretation: Cascade Classifiers and Active Shape Models (ASM). In the first, it uses a classifier capable of detecting the closure of the eyes and in the second the marriage between the ASM response and the Pearson correlation coefficient. The tests show that the model has excellent accuracy and precision and a robust performance in the detection of closed eyes and pose estimation, bypassing very well the natural problems of pattern recognition such as occlusion and illumination. The model response achieved 98 \% average hit with a false positive rate in the house of 2 \%.