A Smart Context-Aware Hazard Attention System to Help People with Peripheral Vision Loss

Area of Science:

• Computer Science
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Peripheral vision loss impairs hazard detection and safe navigation for visually impaired individuals.
• Existing assistive navigation systems often lack hazard prioritization, potentially missing critical dangers.
• Smart glasses offer a platform for wearable assistive technology to augment visual perception.

Purpose of the Study:

• To develop a context-aware, hybrid (indoor/outdoor) hazard classification system for individuals with peripheral vision loss.
• To augment users' healthy vision with smart notifications for potential obstructions in their peripheral field of view.
• To improve navigation safety and cognitive perception for the visually impaired.

Main Methods:

• Utilized computer-enabled smart glasses with a wide-angle camera for real-time object detection using a deep learning object detector.
• Implemented a Kalman Filter multi-object tracker to determine object motion models and extract motion features.
• Employed a neural network-based classifier to categorize detected objects into five predefined hazard classes based on motion features.

Main Results:

• Achieved promising classification performance with up to a 90% True Positive Rate (TPR), 7% False Positive Rate (FPR), and 13% False Negative Rate (FNR).
• Demonstrated an average testing Mean Square Error (MSE) of 8.8% for hazard classification.
• Participant feedback on the system and notification generation was collected, indicating potential user acceptance.

Conclusions:

• The proposed context-aware hybrid hazard classification system effectively augments vision for peripheral vision loss patients.
• Smart notifications generated by the system enhance cognitive perception and awareness of potential hazards.
• The technology shows significant potential for improving navigation safety and independence for visually impaired individuals.