A facial recognition system is a computer application to automatically identify or verify a person from a digital image or a frame from a video source.

But despite decades of research into the technology, experts say they have still to overcome some problems with facial identification.

Zheng Nanning, a top artificial intelligence scientist and member of the Chinese Academy of Engineering, said, "A cutting-edge study is looking at facial identification in an uncontrolled environment, for example, outdoor applications, using time-elapsed photos, and recognition from large crowds."

Zheng also said neurophysiologists have shown that in the right temporal lobe of the human brain a specific area exists that is in charge of facial recognition ability. "How to find this powerful mechanism and inspire the development of facial recognition technology will be a valuable and challenging study direction," he said.

Traditionally, a facial database is collected in the laboratory, where illumination and posture of subjects are controlled.

"Scientists are still looking for better solutions to cope with these factors," Zheng said.

Liu Yuehu, a professor specializing in artificial intelligence at Xi'an Jiaotong University, said current technology cannot completely recognize faces after plastic surgery, or in the case of twins.

"These are difficult to recognize even for human eyes, not to mention computers," Liu said.

Liu added that databases on different races are necessary for facial recognition.

"We Chinese may have this feeling that we can easily tell the difference between two Asian people, but for the faces of other people we may find they look alike.

"This is because our brains do not receive many inputs on the images of those people.

"The theory also applies to computer recognition. Only if we have plenty of data on the faces of different races may it be possible for computers to recognize people from different races," he said.

Liu's team at the Institute of Artificial Intelligence and Robotics has collected 33,669 images from 1,247 volunteers, forming one of the largest oriental facial image databases in China.

The database is used by 23 universities and research institutions in China, and 21 abroad.

(China Daily 05/02/2013 page7)