시각장애인을 위한 XR 글래스 기반 보행 장애물 감지 및 실시간 원격 도우미 통합 보행 지원 시스템

초록

This study proposes an XR glasses-based integrated walking assistance system for visually impaired individuals. The system captures first-person video and performs object detection and risk analysis using an on-device AI module, providing voice guidance via TTS. A WebRTC-based framework enables real-time communication with a remote helper for handling unexpected situations. The system integrates perception and remote assistance to support safe walking in complex environments. YOLOv8-Nano was selected as the optimal lightweight model for real-time performance in mobile environments. Experimental results show a processing time of 0.16 seconds, 95.4% detection accuracy, and 0.04-second communication latency. User evaluation indicates high satisfaction in walking safety and remote assistance, with overall usability above 4.0.

키워드