AI-powered BlindSpot VisionGuide system on raspberry Pi for enhancing independence of visually impaired users

Area of Science:

• Computer Science
• Artificial Intelligence
• Assistive Technology

Background:

• Visually impaired individuals face significant challenges in daily tasks and independent living.
• Existing assistive technologies often lack integration, real-time interaction, or affordability.

Purpose of the Study:

• To develop and evaluate BlindSpot-VisionGuide, an integrated, AI-based assistive system for the visually impaired.
• To empower visually impaired users with real-time audio feedback for enhanced independence.

Main Methods:

• Integrated three core AI capabilities: face recognition (deep facial embeddings), image captioning (BLIP model), and online newspaper reading (API-driven text-to-speech).
• Developed a centralized voice interface for seamless interaction with all system modules.
• Deployed the system on Raspberry Pi hardware, optimizing for offline use and low-power consumption.

Main Results:

• The BlindSpot-VisionGuide system demonstrated reliable performance across all modules on a Raspberry Pi 5.
• Experiments confirmed effective recognition accuracy, acceptable response times, and manageable memory consumption.
• The system successfully balanced computational demands with user-friendliness for practical application.

Conclusions:

• BlindSpot-VisionGuide showcases the practical application of embedded AI in creating scalable and inclusive assistive technology.
• The system provides a foundational model for next-generation vision-based accessibility solutions.
• Potential extensions include object detection, multi-language support, and caregiver integration for further enhancement.