A new study by the University of East Anglia (UEA) suggests computers are now better at lip-reading than humans.
The performance of a computer based lip-reading system was compared to that of 19 human lip-readers. Results showed the computerised system was over 50% better at recognition than the humans completing the same task.
Simultaneously, the study found that while lip-reading humans required a full video of the individual speaking, computers could work with simple features such as the shape of the face.
The study also found that compared to the typical approach to lip-reading training in which individuals are taught to spot lip-shapes from drawn images, the full appearance of speech gestures is very important.
Using this video-based training system, viewers with limited skills improved their ability to lip-read monosyllabic words. This research may improve methods of lip-reading training for deaf and hard of hearing people.
“This pilot study is the first time an automated lip-reading system has been benchmarked against human lip-readers and the results are perhaps surprising”, said the study’s lead researcher, Sarah Hilder.
“With just four hours of training it improved their lip-reading skills markedly. We hope this research is a technological advance for the deaf community.”
RNID campaign manager Agnes Hoctor commented, “We would welcome the development of video-based or online training resources to supplement the teaching of lip-reading”.
(compiled by Miriam Walsh)