How could Santa know if you've been good or bad?
(麻豆淫院Org.com) -- By using technology to detect guilty expressions, of course. CSIRO is using automated expression recognition technology to tell whether someone is in pain and, according to computer scientist, CSIRO鈥檚 Dr Simon Lucey, there鈥檚 no reason why Santa couldn鈥檛 train the system to find out who鈥檚 been naughty or nice.
鈥淓ach facial expression is made up of many different components - a twitch of the mouth here, a widening of the eyes there - some lasting only a fraction of a second,鈥� Dr Lucey said.
鈥淥耻谤 computer program looks at these components, matches them against a list drawn up by expert psychologists and decides what expression just flitted across a face.鈥�
It turns out that most human expressions are the same regardless of background.
And, as anyone who has been naughty knows, it鈥檚 also very hard to fake an expression.
While a guilty person might fool a human with their look of pure innocence, it鈥檚 very hard to fool Dr Lucey鈥檚 computer.
鈥淭here are always some micro expressions that we are unable to control and some of these are associated with deception. It is these tiny facial expression components that the computer can spot.鈥�
The system uses a technique called machine learning. After being programmed with what to look for and what it means, each observation taken or analysis performed is used to refine the computer鈥檚 technique so it gets better at its assigned task over time.
Dr Lucey鈥檚 system uses a webcam and will work against any background, providing there鈥檚 enough light.
鈥淥耻谤 expression work is still mostly from front-on, but we鈥檙e teaching the system to do what we do and recognise expressions when only one side of the face is visible.
鈥淲hen it comes to finding out who鈥檚 been naughty or nice, we show the computer what expressions are associated with good behaviour and it watches for a departure from that,鈥� Dr Lucey said.
One of the applications for Dr Lucey鈥檚 research is in detecting whether someone who cannot communicate is in pain and how intense that pain is.
Another is in making 鈥榯elecollaboration鈥� between people using video and computers in different locations more natural by recognising gestures such as someone pointing at an object.
Provided by CSIRO
