Standalone Smart Glass System for the Blind and Visually Impaired

Blindness stands as one of the most prevalent and impactful disabilities, affecting millions of individuals globally. The present work focuses on developing a smart glass system meticulously designed to cater to the unique needs of the blind and visually impaired (BVI) community. Unlike existing solutions, the proposed smart glass operates as a standalone device, eliminating the need for constant internet connectivity or reliance on external devices such as smartphones or external processor for computational resources. The system comprises three key modules: object recognition, text recognition and face recognition, employing YOLOV8, KerasOCR and FaceNet, respectively. The hardware integrates a Raspberry Pi Zero 2 W and Intel Neural Compute Stick 2, enabling efficient edge-based deep learning model execution. The YOLOV8 model is trained on ImageNet and a custom dataset for Malaysian currency, featuring 270 object classes. To enhance user interaction, offline speech recognition via Vosk-API is implemented, providing input and output through a build-in speaker and MEMS microphone. This multifunctional, lightweight and standalone smart glass system aims to empower the BVI community, fostering self-reliance and emphasising ability over disability.