Crosswalk Guidance System for the Blind

Abstract

Street crossing can be a significant challenge for visually impaired people, limiting their mobility especially in urban environments. To date, there are few solutions for this significant problem. Current approaches for guiding blind pedestrians in crosswalks have mainly focused on detection of crosswalks and crosswalk signals. Few studies have taken into consideration the mobility of a visually impaired person while street crossing. We programmed a commercially available, wearable goggle system to detect crosswalk signals, to plan a path across the street, and to provide verbal guiding cues with real-time semantic features to keep the user on the correct path. During verification testing, we found crosswalk signal detections were typically reliable but depended on hyper-parameters to reduce false positive errors in the crosswalk signs in a small number of cases. Testing with visually impaired subjects resulted in successful guidance at an outdoor crosswalk.