A real-time visual interpretation system for the visually impaired using ESP32-CAM and cloud-based AI Chat-GPT 4 O mini

Huisa A.E.A.; Quispe A.R.R.; Victorio R.D. · 2025 13th International Conference on Control, Mechatronics and Automation, ICCMA 2025 (2025) · DOI: 10.1109/iccma67641.2025.11369696

Resumen

Millions of people around the world face daily challenges due to visual impairments, particularly in navigation and object recognition. While existing tools like canes and guide dogs offer partial support, they often lack real-time context awareness and are inaccessible to many due to high costs or limited availability. This paper presents Vogui, a low-cost, user-centric assistive technology system that integrates an ESP32-CAM module with cloud-based artificial intelligence to deliver real-time auditory feedback. The system captures environmental images, sends them to AI servers for object and pattern recognition, translates the scene into natural language descriptions, and communicates this information to users through synthesized speech. Extensive field testing was conducted in indoor and outdoor settings under various lighting and network conditions (3G/4G). The prototype achieved an average response latency of 1.5 seconds (4G), a recognition accuracy of 90%, and maintained stable connectivity with minimal power consumption. Comparisons against smartphone-based alternatives highlight Vogui's superior specialization, autonomy, and affordability. These findings demonstrate the feasibility and effectiveness of Vogui in promoting independent mobility and social inclusion for the visually impaired, especially in resource-constrained environments.

Instituciones: Universidad Continental