lip reading
Latest
AI is already better at lip reading than we are
Humans are have never been particularly adept at lip reading, but that hasn't stopped us from training AI to do it with nearly 100 percent accuracy.
Lip reading mobiles are wunderbar, still at the prototype stage (video)
We came across this lip reading prototype during our exploration of the CeBIT 2010 halls, and while we're a bit tardy in bringing it to your attention, there's a certain timeless quality to strapping your face with wired sensors that transcends conventional restrictions of timeliness. That's our story anyway. Devised by researchers at the Karlsruhe Institute of Technology, it picks up the motion of speech (via electromyography) without requiring the sound, and then translates it into audible communication via a delightfully cold and robotic voice. The purposes of such a project are obvious -- from helping people who've lost their speech to making private telephone conversations actually private -- but the fun is in seeing someone use the thing in its current unrefined form. You'll be able to do that just past the break.
Lip-reading software can identify multiple languages, has big plans for Jupiter
Some scientists at the University of East Anglia in England, with aims only for the betterment of humanity, we're sure, have developed lip-reading software that can not only pick up words based on mouth shapes, but can even figure out what language you're speaking -- always a handy thing when understanding speech, we've found. Languages it can identify include English, French, German, Arabic, Mandarin, Cantonese, Italian, Polish, and Russian, and recognition is based on telltale "articulators" of tongue, jaw and lip movement. The software statistically analyzed the face movements of bilingual and trilingual speakers to draw out the specific defining characteristics, and could by used by the deaf, military and police organizations, or by on-board ship computers that fear for their safety and the integrity of the mission.[Via CNET]
Crime fighting system converts lip motions into spoken word
If you think text-to-speech or video-texting is hot stuff, researchers at the University of East Anglia have something that just might cool your excitement. Kicking off a three-year project, the team is setting out to "collect data for lip-reading and use it to create machines that automatically convert videos of lip-motions into text." By building a vast database of lip movements that can be read and spat out in verbally, the gurus hope to fight crime by being able to pick out potentially threatening phrases that security cameras pick up. The university is teaming up with the Centre for Vision, Speech, and Signal Processing at Surrey University in order to get the technology wrapped into cameras just about everywhere, from mobile phones to vehicle dash boards. Admittedly, we're not exactly keen on yet another Big Brother agenda gaining traction, but if this stuff stays in the right hands, you loud-mouthed criminals better start crafting your own unique language over the next few years.
