Automated Book Reader for Persons with Blindness

摘要

This research introduces a new automatic book reader for persons with blindness. The objective is to design a fully integrated system that is relatively fast and yet inexpensive and effective with a high reading accuracy. Through the use of two inexpensive light weight cameras, a book holder and using regular lighting of an office or a lab environment, this integrated system addresses through software development (a) the mathematical foundation of perspective distortion introduced by page curvature of an open book, and of barrel correction introduced by the inherent nature of image capture of the camera; (b) image preprocessing for finding lines and characters in a given image; and (c) the implementation of a fast neural network that takes as input the findings of step (b) and provides an audible read out through a speech synthesis engine.