Facing up to the computer age
Logging on to inappropriate websites by under-age computer users may be about to become a little more difficult than merely pressing the “I CONFIRM” button.
Using
mathematical algorithms, Deakin researchers Professor Kate Smith-Miles
and PhD student Xin Geng are working on an automatic age estimation project
that will help determine whether the face of the person at the keyboard
conforms with the age they say they are.
“That’s just one practical and obvious way in which the work we’re doing could be used,” Professor Smith-Miles said of the work which has already gained extensive global recognition.
A paper co-authored by Professor Smith-Miles and Xin Geng with Professor Zhou Zhi-Hua from China’s Nanjing University has been published as the feature article in the December edition of the prestigious American-based journal – IEEE Transaction on Pattern Analysis and Machine Intelligence.
The Deakin researchers also identified four other practical applications for their work in the paper:
1. Age-specific human-computer interaction: People at different ages have different requirements and preferences for interaction with computers, including linguistics, aesthetics and consumption habits. If computers could estimate the person’s age, they could automatically choose the vocabulary, interface and services suitable to the user.
2. Multi-cue identification/verification: Although age is not a reliable biometric feature for identification and verification, automatic age estimation can work together with other widely used biometric trains like fingerprints and iris recognition to improve security.
3. Law enforcement: Age is always an important attribute when describing a person involved in law enforcement procedures. Sometimes the age of a person in a photo is not known. The technique of automatic age estimation could help police determine the age of a suspect more accurately and efficiently.
4. Understanding the aging process in other areas: The research on automatic age estimation algorithms could provide valuable help to researchers in psychology, medicine and other fields about the aging procedure or the perception of aging variation.
“While recognition of most facial variations, such as identity, expression and gender, has been extensively studied, automatic age estimation has rarely been explored,” said Professor Smith-Miles, the head of Deakin’s School of Engineering and IT.
“In contrast to other facial variations, aging variation presents several unique characteristics which make age estimation a challenging task. In this paper we proposed an automatic age estimation method named AGES (AGing pattErn Subspace).
“The basic idea is to model the aging pattern, which is defined as the sequence of a particular individual's face images sorted in time order, by constructing a representative subspace.
“The proper aging pattern for a previously unseen face image is determined by the projection in the subspace that can reconstruct the face image with minimum reconstruction error, while the position of the face image in that aging pattern will then indicate its age.
“In our extensive experiments of over 2000 faces, our method outperformed the existing published approaches, and even outperformed human perception of age estimates when the humans were given only the same tightly cropped face images to view as those fed into our algorithm.
“When humans are given wider shots of faces, including hair and clothing, their ability to estimate age is much improved, but without those extra cues our algorithm performs better than humans at age estimation.”
“We have also applied a similar approach to the verification of identity from facial images, and proposed new methods for fusion of biometrics such as facial analysis and gait analysis for the identification of individuals from a database. These applications of computer vision will become increasingly important for national security.”
For more information about the School of Engineering and IT: http://www.deakin.edu.au/scitech/eit/index.php