BI Intelligence
This story was delivered to BI Intelligence "Digital Media Briefing" subscribers. To learn more and subscribe, please click here.
Google’s DeepMind and the University of Oxford are working on a lip-reading system powered by artificial intelligence, New Scientist reports.
The AI system already outperforms professional lip-readers by a mile, opening doors to exciting new opportunities in consumer technology.
The two organizations applied a deep learning system to a large dataset of BBC programs. In total, the AI was trained with 118,000 sentences, using 5,000 hours of video from six TV shows that aired between January 2010 and December 2015. The system was then tested on live broadcasts between March and September 2016.
In a controlled test, the AI blew away professional (human) lip readers. Tasked with transcribing 200 randomly selected clips from the dataset, the professional correctly annotated just 12.4% of the words, compared to the AI which got 46.8% of all words correct. This AI system is also said to be more accurate than all other automated lip-reading systems.
This system is relevant to any context that uses speech recognition and a camera, such as:
To receive stories like this one directly to your inbox every morning, sign up for the Digital Media Briefing newsletter. Click here to learn more about how you can gain risk-free access today.