VIS2SOUND on Reconfigurable Hardware

This work describes the implementation of a bio-inspired visual processing system on configurable logic, designed for the enhancement of relevant information on a real scene, for its use on a complete system to assist the visually-impaired. For this purpose, a sensorial transduction module has been developed, which transforms visual information into sound patterns with 3D spatial localization. The requirements for this application, as portability, constrained power consumption, real-time performance, and customization to fit the specific features of the patientpsilas affection, lead us to consider programmable logic as the best option for the design and development of the whole system, specifically FPGAs. A prototype of the system, including two video inputs with three spatial filters, two weighed sums, a disparity computation module, and the 3D sound generator, has been implemented on a Virtex-II XC2V6000, and performs at a maximum rate of 60 frames per second.