Simulación de un sistema de control de una silla de ruedas basado en visión artificial mediante movimientos de cabeza, para personas cuadripléjicas.

Abstract

People with quadriplegia depend on another person to move, by the lack of mobility in their 4 limbs, using manual wheelchairs. To use motorized wheelchairs gives them some independence, needing methods to control them. For example, theses have been developed controlling motors by voice or head movements detected by accelerometers. The latter uses a headband with the chip that contains the accelerometers, which can be annoying for the user. The present work seeks to be less invasive, for this reason a system for detecting movements of the head has been developed through artificial vision avoiding placing any device and detecting these movements through a strategically located camera. The system starts by detecting the face with the Viola-Jones algorithm, then the section of the nose is extracted for monitoring using the Kanade-Lucas-Tomasi algorithm and the new regions of the face and nose are obtained with the geometric estimate. To enter/exit command mode, user turns head from left to right in less than 3 seconds maximum. As long as this does not happen, the user can move his head freely without activating the motors. Finally, if the user enters the system and tilts his head to the left, right, forward or backward the motors will be activated in the detected direction. This prototype, due to the pandemic situation, has only been simulated and does not really interact with a wheelchair, which is not a problem since this part has been solved in those previous theses.