An Artificial Intelligence Edge Computing-Based Assistive System for Visually Impaired Pedestrian Safety at Zebra Crossings

This article proposes a wearable assistive system based on artificial intelligence (AI) edge computing techniques to help visually impaired consumers safely use marked crosswalks, or zebra crossings. The proposed wearable assistive system consists of a pair of smart sunglasses, a waist-mounted intel...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on consumer electronics 2021-02, Vol.67 (1), p.3-11
Main Authors: Chang, Wan-Jung, Chen, Liang-Bi, Sie, Cheng-You, Yang, Ching-Hsiang
Format: Article
Language:English
Subjects:
Artificial intelligence
Artificial intelligence of the Internet of Things (AIoT)
Collision avoidance
Consumers
Dogs
Edge computing
Global Positioning System
Legged locomotion
Object recognition
Pedestrian crossings
Pedestrian safety
pedestrian walking safety
Pedestrians
Real time
Safety
Smart glasses
Sunglasses
Traffic safety
Traffic signals
Visual impairment
visually impaired
Walking
walking cane
wearable assistive devices
Wearable technology
zebra crossing
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This article proposes a wearable assistive system based on artificial intelligence (AI) edge computing techniques to help visually impaired consumers safely use marked crosswalks, or zebra crossings. The proposed wearable assistive system consists of a pair of smart sunglasses, a waist-mounted intelligent device, and an intelligent walking cane (stick). A deep learning technique is adopted for zebra crossing image recognition in real time. Visually impaired consumers need to wear the proposed smart sunglasses and waist-mounted intelligent device and hold the proposed intelligent walking cane when they approach a zebra crossing. When a visually impaired pedestrian reaches a zebra crossing, they will immediately receive a message about the current situation at the crossing and the traffic light signal. Experimental results show that the accuracy of real-time zebra crossing recognition of the proposed system can reach up to 90%.
ISSN:0098-3063
1558-4127
DOI:10.1109/TCE.2020.3037065