Visually impaired individuals encounter significant challenges in safe navigation due to limited environmental perception and dynamic obstacles. Existing assistive devices primarily provide proximitybased alerts without contextual understanding. This paper proposes a Vision-Assisted Smart Navigation System that integrates embedded artificial intelligence and multi-sensor fusion to enhance mobility and safety. A lightweight deep learning model performs real-time detection and classification of obstacles such as humans, vehicles, walls, and staircases, while ultrasonic sensors provide accurate distance measurement. A Kalman filterbased fusion approach improves reliability and reduces sensor noise. The system delivers intuitive feedback through voice guidance and haptic alerts, enabling timely and effective user response. Additionally, GPS and IoT connectivity support real-time location tracking and emergency alert functionality. The prototype demonstrates low latency, high detection accuracy, and energy-efficient performance in real-world conditions. The results indicate that the proposed system significantly improves navigation, safety and independence, making it a practical and cost-effective solution for assistive mobility.
Panchetti et al. (Wed,) studied this question.