Assistive Technology for the Visually Impaired Using Computer Vision

India has more than a quarter of the world's 36 million blind people. Educating the blind to avoid unemployment among their population is one of the most challenging problems faced by blind schools today. Although many schools resort to the use of Braille to eradicate illiteracy among them, its steep learning curve, insufficient availability and high cost makes it quite unapproachable. Braille Literacy Statistics in India indicate that out of the 12 million blind people in the country, less than 10 percent learn Braille. It is evident that one of the most crucial issues faced by the blind and the visually impaired is the inability to read and learn without the use of Braille. In order to overcome this issue, there is a need to develop a system that can assist the visually impaired in reading. Hence, the proposed solution is to design an inexpensive wearable device that uses computer vision to read out any form of text around the user in various alignments and lighting conditions. The system makes use of a Raspberry Pi with a compatible camera to capture the content around the visually impaired or blind person and reads it out to them in a regional language. A sensor is also incorporated to notify the user of the distance to the nearest object at his eye level and the device enumerates various objects in its sight. The system is based on the combination of image processing, machine learning and speech synthesis techniques. The observed accuracy combining both the optical character recognition and the object recognition algorithms was found to be 84%.