A computer scientist from Umeå University in Sweden has developed an algorithm that can recognise a face from just one picture kept in a database. The system can also determine any likeness you may share with another person, such as a celebrity.
Developed by Hung-Son Le as part of his PhD dissertation, the system, which takes into account light conditions and facial expressions, uses a Hidden Markov Model (HMM) negating the need for time to ‘know’ new pictures with different expressions taken under different lighting conditions.
To test the face recognition system, it was set against international standard systems such as FERET and the Yale Database – Le’s system outperformed its competitors.
As part of the many commercial applications possible based on the technology, a face search engine is in a final development phase. It will soon be accessible via links published in Hung-Son Le’s homepage.
You’ve read it. Now review it.
Date Published: January 28, 2008
More by this source
|
Print
|
Send to a friend
|
Rate & Comment
|
Keep up to date
If you found this item fun or informative, please let others know. Simply send to a friend or recommend it to even more people - on any of the following sites:
Latest Science News | reddit | digg.com | del.icio.us | rollyo | stumbleupon
More on facial recognition...
Facial analysis
German researchers have developed software that can identify human faces in an image and then analyse their expressions.
Has voice recognition finally come of age?
Speech recognition - enabling computers to transcribe the human voice into readable words - has been tomorrow's technology for so long you could be forgiven for thinking it would never come. And there has always been the problem of how useful it would be when it did arrive if it was anything less than perfect.
No hiding place
New 3D imaging technique could lead to crime-stopping facial recognition.
