Assisting the visually impaired: obstacle detection and warning system by acoustic feedback

Area of Science:

• Computer Vision
• Assistive Technology
• Human-Computer Interaction

Background:

• Visually impaired individuals face significant challenges navigating environments.
• Existing assistive technologies often lack comprehensive real-time obstacle detection capabilities.
• The need for intuitive and non-intrusive sensory feedback is crucial for effective navigation aids.

Purpose of the Study:

• To design and evaluate an obstacle detection system for the visually impaired.
• To leverage stereo vision for 3D obstacle perception in diverse environments.
• To integrate a user-friendly acoustic feedback mechanism for enhanced safety.

Main Methods:

• A stereo camera system captures visual data.
• Dense disparity map computation for 3D scene reconstruction.
• RANSAC-based ground plane estimation for robust environmental understanding.
• Polar grid representation for obstacle mapping.
• Bone-conducting audio technology for non-intrusive acoustic feedback.

Main Results:

• The system successfully detects potential obstacles in 3D across indoor and outdoor settings.
• Robust ground plane estimation was achieved using RANSAC and filtering.
• Acoustic feedback effectively alerted users to nearby obstacles.
• User study with visually impaired volunteers indicated system efficacy.

Conclusions:

• The proposed obstacle detection system enhances navigation safety for visually impaired users.
• The integration of stereo vision and acoustic feedback offers a promising assistive solution.
• Further development could refine obstacle classification and user customization options.