VisionAssist: A Voice-Based Assistive System for the Visually Impaired

Visually impaired people find it difficult to know who or what is in front of them, and they depend on other people for navigation, reading and daily tasks. Existing aids such as the white cane detect only obstacles within contact range, smartphone applications are not hands-free and usually need an internet connection, and smart glasses are costly and heavy. This paper presents VisionAssist, a low-cost, wearable and voice-based assistive system. The user presses a single push button to capture an image with an ESP32-CAM, while an HC-SR04 ultrasonic sensor measures the distance to obstacles ahead. The ESP32 sends the image and the distance reading to a backend server over Wi-Fi. The backend identifies people, animals and objects using a deep-learning object detector, reads printed text using optical character recognition (OCR), checks the obstacle distance, and converts the result into speech using a text-to-speech (TTS) engine, which is played through a speaker or earphone. The system deliberately avoids heavy scene description so that the spoken output stays short, fast and useful.