University:
California State University, East Bay
Course:
CS 3590 | Data Communications and Networking
Academic year:

2023
Views:

283

Pages:

2
Author:

Jithin J. I.

AI-BASED ASSISTIVE SYSTEMS FOR VISUALLY IMPAIRED PEOPLE Introduction  Visually impaired people face difficulties in mobility and risk of accidents  White cane and guide dogs have limitations in detecting aerial obstacles  Camera-based solutions can provide enhanced environment perception Vision-Based Assistive Systems  Use cameras and computer vision for obstacle detection and navigation  Image processing algorithms analyze video frames  Provide feedback to user via audio, haptics etc.  Solutions using color markers, QR codes, RFID for indoor navigation  Outdoors, combine with GPS, GIS database for wayfinding  Object recognition to detect landmarks like doors, signs  Deep learning models being explored recently for robust performance Challenges  Accuracy in estimating distance to obstacles  Scale drift between different scenes  Handling occlusions, cluttered environments  Real-time performance with low power consumption Conclusion  AI and computer vision have huge potential for assistive technology  Robust perception, efficient models and intuitive interfaces needed  Enhanced environment understanding with minimal training data  Multi-modal interfaces for customizable user experience Key Highlights  Camera-based solutions are promising for visually impaired assistance  Solutions exist for indoor navigation, outdoor wayfinding, object detection  Deep learning models being leveraged recently for accuracy and robustness  Key challenges include real-time performance, scalability, and usability  AI and computer vision will play a big role in future assistive technologies