Horizon Cyber-Vision: A Cybernetic Approach for a Cortical Visual Prosthesis

Abstract

The way towards the next generation of visual cortical prosthesis is visualised through an engineering cycle based on a cybernetic paradigm. Our proposal is to develop a configurable and wearable system that will generate simulated prosthetic vision, while on the other hand, perform intracortical stimulation when applied to blind patients, so that it is expected that improvements with sighted volunteers, in combination with transformed reality strategies, will correlate with similar improvements in blind patients. The resulting cybernetic model involves modelling from stimuli to visual percepts, and in parallel, developing the best suited transformed reality strategy leading to a better perception of the environment. Deep learning approaches for object detection, monocular depth estimation, or structural edge detection, in combination with the use of an eye-tracking system, will lead to an integrated system that has proved to be wearable, optimised, modular, and computationally lightweight. To assess the cybernetic approach, behavioural experiments are proposed using two different scenarios. Firstly, a corridor with a series of obstacles and a controlled but more complex environment that resembles a city square, called StreetLab.