A smile consists of two differentiating periods of motions of mouth - neutral to smile and smile to neutral. Ning et al. conducted a study to determine discriminating power of a smile based on these two periods of motion.
Landers et al. proved that people are more likely to recognize faces from real smiles than fake smiles as well as from moving images than static images of smiles. Using this conclusion, Ning et. al., prepared a video database of 30 to 40 video clips of 10 subjects. Each subject performed smile expression as naturally as they could. The expression begins with a neutral face, goes to smile, and from smile to neutral face again.
After examining the features generated from neutral to smile and smile to neutral periods of motions separately, they found that both give the same information about identity.
This implies that the discriminating power of smile based on neutral to smile and smile to neutral periods of motion is same. Motion doesn’t matter to people; what matters is whether they go flat on charm in your smile when your expression goes from neutral to smile.
A human face contains a lot of information about identity. In further investigation, they tried to know which part of face is more discriminating. As features were generated from smile dynamics, error rates of lower face and upper face were compared separately.
They found that lower face is 3 times more discriminating than upper face. So, if you want people to remember you for a long time, smile more.
-------------------------------------------------------------------------------------------------------------------------------------------
Reference - Ye Ning, Terence Sim. School of Computing, National University of Singapore. Research paper name - Smile, You’re on Identity Camera. Published in - IEEE (2010)
Reference – K. Lander, L. Chuang and L. Wickham. Research paper name – Recognizing face identity from natural and morphed smiles. Published in – The Quarterly Journal of Experimental Psychology (2006)
-------------------------------------------------------------------------------------------------------------------------------------------
(Image courtesy of stockimages at FreeDigitalPhotos.net)
No comments:
Post a Comment