A System Design for an AI Based Voice Assistance System with Real Time Output for Glaucoma Patients

Glaucoma causes progressive periph- eral vision loss, severely limiting environmental awareness and increasing the risk of accidents during daily navigation. Existing assistive tools are often limited, expensive, or focused on a single function, while current AI-based solutions remain fragmented across object detection, text reading, or scene understanding. This paper presents a unified AI-Based Assistive Vision System for Glaucoma Patients, designed to compensate for peripheral field loss through intelligent auditory guidance. The system integrates YOLOv5 for real-time hazard detection, OCR for reading environmental text, and a transformer-based image captioning model for scene-level semantic interpretation. A fusion layer prioritizes safety-critical information and produces concise audio descriptions through text-to-speech, enabling low-latency, portable deployment on edge devices. By combining multi-modal perception with real-time summarization, the proposed architecture offers a scalable and robust solution that enhances mobility, situational awareness, and independence for individuals with glaucoma.