Read my Lips

Lip reading technology just got better. Credit: University of East Anglia

Visual speech recognition is a branch of visual recognition focussing on understanding speech by observing facial expression with a particular focus on reading lip movements.

We, as human beings (but we are not alone, dogs are pretty good at evaluating our facial expressions), make use of visual support in understanding what another person is saying. This usually complement the sound (and you can appreciate this directly by noticing how easier it is to understand a person speaking a language that is not our native tongue by looking at his face than just hearing him talking over a phone) but is some cases it can actually substitute the sound when you are hearing impaired.

Having a computer reading the lips of a person to understand what he is actually saying is quite complex and requires some sort of intelligence. 

As a matter of fact you need to recognise the way lips move and associate to each sound a specific lip movement. Unfortunately, there are a few sounds that look exactly the same, like the pronunciation of "p" and "b" (look at your lips in a mirror and see for yourself).  Hence a computer has to find out what would make sense, something our brain is very good at (to the point that it usually hears what it makes sense to hear: the brain automatically replaces aural and visual clues with those that make sense).

Now researchers at the University of East Anglia have been able to create a software that is pretty good at reading lips.

The improvement is based on a software that trains the computer to distinguish among different movements of the lips and put these movements in the context of a "meaningful" speech.

This requires quite a bit of artificial intelligence and it is nice to see the progress being made in this area.

An automatic lip reading system can find several applications, from assisting hearing impaired people to capturing speech through surveillance cameras for security reasons (surveillance cameras typically o not capture sounds).

On the other hand lips reading gives another blow to our desire for privacy.... There is also another face in the technology coin!

Author - Roberto Saracco

© 2010-2020 EIT Digital IVZW. All rights reserved. Legal notice. Privacy Policy.

EIT Digital supported by the EIT