Maths enables computer to identify gender

Computers can mimic human perception of gender, according to a multi-disciplinary team of three computer scientists and two human anatomy experts at the University of Western Australia.

For the first time, the team developed a mathematical model where a computer could be used rather than human ‘raters’ to match the gender scores they give to human faces that range on a continuum from very masculine to very feminine.

“A subjective gender score is a tangible metric that human raters assign to the degree of masculinity-femininity of a face,” the researchers write in a report on the study.

“This is because, though sex is binary, gender is understood to be a continuum. For example, the illustration shows synthetic images of the same individual by varying its gender from very male to very female.”

The researchers note that gender scores have long been used in psychological studies to understand the complex psychosocial relationships between people while perceptual scores for gender and attractiveness have been employed for quality assessment and planning of cosmetic facial surgery.

Similarly, they say that various neurological disorders have been linked to the facial structure in general and the facial gender perception in particular.

“While, subjective gender scoring by human raters has been a tool of choice for psychological studies for many years, the process is both time and resource consuming,” the researchers write in their report, published in the journal PLOS ONE.

“In this study, we investigate the geometric features used by the human cognitive system in perceiving the degree of masculinity/femininity of a 3D face.”

For the experiments, the team obtained 3D face scans of 64 students from the university – 34 female and 30 male – using a 3dMDface scanner.

Because skin texture itself can be used to identify gender, the 3D images were made ‘textureless’ to avoid any bias in the results. Likewise, because hair is one of the major contributors in sex classifications, ratings were obtained on the images with the hair concealed or cropped.

“The textureless 3D face scans of the subjects were then observed in different poses and assigned a gender score by 75 raters of a similar background,” the report states. “Our results suggest that the human cognitive system employs a combination of Euclidean and geodesic distances between biologically significant landmarks of the face for gender scoring.”

Lead author and PhD candidate Syed Zulqarnain Gilani said the model would be useful in quickly and accurately evaluating gender scores in research, such as investigating the relationship between masculinity and femininity and health, and in evaluating cosmetic facial surgery in terms of attractiveness, pre- and post-operation.

“Until now the tool-of-choice for getting a gender score has been to call in subjects – sometimes as many as 300 per study – and to recruit ‘raters’ to give each subject’s face a score,” Gilani said.

“Sometimes almost 700 raters might be needed for a study, giving as many as 22,000 ratings which then have to be evaluated. This is a very cumbersome and slow process.”

The mathematical model used by a computer, however, is able to automatically assign an objective gender score to a 3D face with a correlation of up to 0.895 with the human subjective scores. Such a system could be used to replace the human raters that are usually required.

Syed’s next project, a logical step from this, is to work with colleagues at the university, the Telethon Institute for Child Health Research and the Princess Margaret Hospital in Perth to analyse the 3D faces of children to see if autism, believed to be related to testosterone levels, can be detected and treated early.

* Photo: The photo shows morphed 3D images of the same individual with gender varying from highly masculine to highly feminine.