A New Lip-Reading Computer Has Been Developed By Scientists That Have Been Proven To Be More Accurate Than Humans.

Google's artificial intelligence DeepMind division funded the Lipnet team from Oxford University to build this device. Scientists have developed unique computer software that can read a person's lips with an advanced level of accuracy that is higher than trained experts. This program which is known as Lipnet was developed by researchers at the University of Oxford. This program can accurately read lips 93.4% of the time.

"Machine lip readers have enormous practical potential," explained the team in a paper outlining the invention.

They say the uses include: "Applications in improved hearing aids, silent dictation in public spaces, covert conversations, and speech recognition in noisy environments, biometric identification, and silent-movie processing."

There are over 30,000 three-second videos used to train the computer with people speaking sentences. This enabled the program to learn how to match the different lip movement with the words that are being spoken. Researchers have tested Lipnet on three experts from the Oxford Student's Disability Community, with surprising results. While the experts had an average error rate of 47.7%, LipNet was wrong only 6.6% of the time. The videos that were used to learn to lip-read all followed a certain type of syntax - it would need to study many more to be able to understand everyday speech. This amazing software is not ready to be released to the world as yet.

Yannis Assael, an author of the project contacted Mirror Online to allay has concerns that the program could be used for mass surveillance.

"LipNet has no application in the world of surveillance, simply because lip reading requires you to see the subject’s tongue - meaning that the video has to be straight on and well-lit to get a good result (as we analyse in our paper)," he said.

"It’s technically impossible, to use any lip-reading software for surveillance since the frame rate and the resolution are also factors; and something that is usually neglected with CCTV. If one has a frontal video of someone taken with a very good camera, then you probably have a directional microphone (pointed at them) as well. Hence, we can focus on the positive sides of such a project."