Lip-reading software developed by researchers at the University of California, Merced (UC Merced)'s Human-Computer Interaction Group can continuously learn from its users.
LipType allows users to send texts or emails on their computers and mobile devices, and contactlessly engage with public devices without speaking aloud.
The UC Merced team incorporated filters for different lighting conditions and a mistake-corrector based on different language models into LipType, which proved faster and less error-prone than other lip-reader software.
UC Merced's Laxmi Pandey said, "LipType performed 58% faster and 53% more accurately than other state-of-the-art models in various real-world settings, including poor lighting and busy market situations."
From University of California News
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found