Object detection and conversion of text to speech for visually impaired

Abstract

Assistive technologies are being developed for visually impaired people in order to live confidently[1]. In this project work, we aim to develop a system which would help blind persons get information about objects present in their surroundings in their daily lives. The project work is framed into two stages.  First, image is captured using a portable camera module, if the object is identified as cell phone, person, book or as such, then the detected object is matched with a predefined dataset. A predefined dataset is loaded in order to match the detected object with the captured image. Secondly, once it is matched the recognized text is synthesized for producing speech output. Text to speech conversion successfully converts the detected object into an audio signal using the gTTs Module with the help of the iPython audio library.  The objective of the TTS is the conversion of text into a natural language.  It is not only applicable for the visually impaired but also to any normal human beings who    are willing to read the text as a speech as quickly as possible. The entire system is de-signed using the YOLOv4 model trained on the MS COCO dataset through a laptop. The entire system is processed by using Python as the programming language