Desarrollo de un prototipo de dispositivo para reconocimiento de texto tipo imprenta y su conversión en audio orientado a personas con discapacidad visual mediante técnicas de visión artificial.

Abstract

The present graduate research project was to develop a prototype for recognizing type text printing press with a voice synthesizer aimed at people with visual disabilities using artificial vision. The prototype was built using a single-board computer called Raspberry Pi 3B+, easily accessible, with a free Linux software called Raspbian. The Python software will be installed on Raspbian to develop the program that consists of recognizing characters starting from the capture of a text image through the algorithm OCR Tesseract. The algorithm will extract the text from the image later through GTTS (Google Text to Speech) convert it to audio in a speech to access the information. In respect of hardware, a structure capable of containing the electronic elements was designed that are part of the device such as Raspberry pi 3B+, Raspberry Pi camera, Arduino nano development board, speakers, limit switch sensor, switch, push-button, LCD, and a led panel to be in charge of lighting the system. Functional tests were carried out with all three types of fonts most used, Calibri, Times New Roman, and Arial, together with the correct lighting and the range of distance the camera should be from the sheet of text for good efficiency. Thus, with illumination greater than 180 lumens, at a distance of 25 ± 5 cm from the camera, Calibri font with a size of 12 pts., in normal format with a text of 150 ± 50 words, has an efficiency of more than 95%. It is recommended to carry out offline learning called trainer, making the algorithm can learn by itself.