An Intelligent Voice Assistance System for Visually Impaired using Deep Learning

Abstract

Unassisted navigation, object recognition, obstacle avoidance, and reading activities are extremely difficult for people who are completely blind. For those who are visually impaired, we present a new form of assistive technology. Raspberry Pi 3 Model B+ was selected to illustrate the proposed prototype's capability because of its inexpensive price, compact size, and ease of integration. Incorporated within the design is a camera, sensors for obstacle avoidance, and powerful image processing algorithms for detecting and classifying objects. Both the camera and the ultrasonic sensors are used to determine the user's distance from the impediment. The image-to-text converter, followed by audio feedback, is integrated into the system. A typical pair of eyeglasses can be used to mount the entire system, which is small, light, and simple to use. Using 60 completely blind people, researchers compare the suggested device to the classic white cane in terms of performance. Controlled environments based on real-world scenarios are used to conduct the evaluations. In comparison to a white cane, the proposed device provides higher accessibility and comfort, as well as simplicity of navigation for visually impaired people.