AI Visual & Reading Aid for the Blind

Abstract

Blindness greatly restricts a person’s ability to move independently, recognize their surroundings, and access written information, making routine activities difficult without external help. The AI-Based Visual Aid with Integrated Reading Assistant is a wearable assistive solution created to support fully blind individuals by merging intelligent navigation functions with real-time reading assistance. Developed using the power-efficient Raspberry Pi 4, the device combines a camera module and ultrasonic sensors to locate objects, estimate distance, and provide voice feedback without requiring user intervention. Using the TensorFlow SSD Lite-MobileNet deep-learning model, the system identifies objects in real time while estimating proximity through a fusion of camera-based interpretation and ultrasonic measurements. To further improve accessibility, the system includes a Tesseract-based OCR module capable of detecting text in captured images and converting it into synthesized speech. The lightweight, hands-free design can be mounted on standard eyeglass frames, allowing comfortable and portable usage. Performance evaluation with 60 fully blind participants in controlled environments showed noticeable improvements in navigation speed and confidence when compared with conventional white-cane use. Participants also highlighted strong obstacle-detection accuracy and reliable reading performance under proper lighting. The modular architecture enables future enhancements such as advanced OCR capabilities, expanded object datasets, slippery-surface alerts, stair recognition, and GPS-assisted guidance. Overall, the project contributes to the development of affordable, intelligent assistive tools that enhance autonomy, safety, and daily accessibility for individuals with visual impairments.