Key points are not available for this paper at this time.
• Custom model trained on personalized dataset for real-time guidance • Wearable device with lightweight frame for comfort and durability • Optimized energy consumption system with Raspberry Pi Zero 2W • Finite element structural integrity analysis and Real-world testing of a developed prototype Physical activity is essential for overall health, yet visually impaired people encounter challenges that highlight the necessity to improve independence and accessibility in this area. In this paper, artificial intelligence (AI) and engineering are integrated to develop an assistive device for visually impaired runners that provides real-time auditory guidance for safety and lane tracking. A machine learning model is trained using a custom-built database to classify safety zones, and an integrated line detection algorithm is used to ensure lane monitoring. The adaptive mechanical design focuses on a lightweight, durable, and ergonomic glasses frame, which incorporates miniature and powerful electronic components that maximize energy efficiency and ensures reliable performance in dynamic conditions. Using a prototype, real-world evaluations are performed that demonstrate accurate detection of the safety zones and lane limits, with correct auditory notifications, proving 92.5% accuracy and 64 ms average end-to-end device latency. The prototype validates the device capability in improving safety and autonomy of visually impaired people, while also demonstrating its affordability, ease of use, and suitability for classic running environments and highlighting the potential of AI-driven assistive technology.
Măgurean et al. (Sun,) studied this question.