Authors:
S. R. Kausiq, A. S. Nawin Raju, K. Aathithiya Balaji, R. Angeline, Prasanna Ranjith Christodoss
Addresses:
Department of Computer Science and Engineering (AI & ML), SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India. Department of Computing, Mathematics and Physics, Messiah University, One University Ave, Mechanicsburg, Pennsylvania, United States of America.
Cloud-based assistive glasses for the visually impaired have major concerns like slow response time, the need for internet wherever they go, and privacy concerns as they send photos to the cloud for the response. This paper shows the complete offline smart glasses prototype with a small camera and bone conduction speakers mounted with a glass, that uses a fully offline model Yolov8n for better efficiency and dual pipeline interaction for the user experience (always on ultrasonic sensor for hazard detection and trigger-based Event for object detection and text-to-speech contents) Offline ensures privacy, low consumption of battery and remote connection everywhere, addressing the critical gaps in the existing model/technologies. Used the Yolov8n pre-trained model for training datasets based on our use cases. The primary use cases of our glasses are urban street crossing (detection of traffic lights, road, moving vehicle at the range of 2-4 meters), indoor navigation for preventing furniture collisions, shopping and daily purchases without assistance, document and sign reading using TTS, workplace independence, minimising colleague dependency, and in emergencies. The glasses aren’t just for object detection; they’re a complete daily companion that handles 90% of navigation challenges that existing glasses fail at.
Keywords: Global Vision Distribution; Current Assistive Wearables; Text to Speech; Event Triggered Pipeline; Artificial Intelligence; Machine Learning; Visible Light Communication; Augmented Reality; Deep Learning.
Received on: 05/06/2025, Revised on: 16/08/2025, Accepted on: 23/09/2025, Published on: 09/03/2026
DOI: 10.69888/FTSBE.2026.000671
FMDB Transactions on Sustainable Biomedical Engineering, 2026 Vol. 1 No. 1, Pages: 33-45