Computers pick-up on body language
By John Lister
A British university has found a way for computers to automatically recognize non-verbal cues such as facial expressions. It could make it easier for machines to detect emotions in users.
The work was inspired by studies from the Engineering and Physical Sciences Research Council into lip-reading, which found that simply being able to detect the words spoken was not as insightful as adding in facial expressions, gestures, the direction people looked and other visual cues.
One of the main differences with this research is that previous studies have involved the human participants intentionally displaying specific emotions to make it easier for the computer to distinguish between them. In the new study, the researchers wanted the participants to act and react naturally, which meant the computers were able to detect subtle, unconscious physical reactions.
To pull this off, the researchers at the University of Surrey filmed people in situations with as little artificial set-up as possible. They then asked other participants to watch the clips and pick out the emotions and meanings that the speakers were conveying beyond their literal words. This data was then fed into the computers so that they could build a link between this communication and the visual cues which could be automatically detected.
The researchers say there are possible uses for the technology both in display and reading by computers. The findings could be used to tweak the way video game characters appear on screen, such that the subtleties of their facial expressions and gestures are more appropriate to their fictional situation and thus more realistic.
The technology could also allow computers or equipment fitted with a camera to detect the emotions of users. That could apply to research programs, such as finding out how users react to a particular menu system, for example on an information screen. It could also be used for real-time information, such as noting when an air traffic controller appears to be showing dangerous signs of stress.
Whether the technology can be used to allow Windows to realize that the verbal message “Oh of course, that’s EXACTLY what I wanted to do when I pressed that button, that’s just brilliant, no really, well done” may lack sincerity remains to be seen.
Related:
Entry was posted
on Tuesday, October 27th, 2009 at 2:47 pm
under
Read more entries by John Lister.
Stumble It!
October 27th, 2009
I’m surprised the phrase “Neural Network” isn’t mentioned anywhere in the article because, as far as I can tell, that’s what this is. I allows a computer to “learn” over time by establishing patterns based on increasing amounts of input (ala speech recognition). It’s a good idea, albeit not all that new. There are many potential uses for this, so I’m looking forward to what other uses people come up with.