These Computers Perceive Gender

Added by Sarah Takushi 

A multi-disciplinary team of scientists working at the University of Western Australia (UWA) has developed a mathematical model that allows a computer to identify the gender of human faces based on a sliding scale. Ranging from the super feminine to the super masculine, faces are objectively assessed by a computer based upon measurements between facial landmarks.


The mathematical model is anticipated to be quite useful for speeding up the research process on sexually linked facial features, as well as lead to future applications in the fields of facial surgeries and diagnostics.
Beyond displays of personalized gender expression such as hair length, jewelry, and/or the use of make-up, humans are adept at analyzing a host of facial features to assess the sex of unfamiliar individuals. Some of these characteristics, such as the presence of facial hair or an Adam’s apple are obvious, but other measures are more subtle. For example, the stereotypical female human face has a narrower chin and rounder corners that give the face a heart-shaped appearance. By contrast a typical male human face has a squarer chin, a higher hairline, and a more pronounced ridge of bone that runs across the forehead directly above the eyes (aka the brow bossing). Some characteristics such as the larger appearance of a woman’s eyes relative to her face or the naturally arched quality of a female’s eyebrows can be further accentuated with cosmetics or personal grooming.
Despite all the measurable differences between prototypical female and male faces, until now the science of assessing the gender quality of faces has had to rely on the subjective opinion of many participants. A single study may require as many as 700 people to participate and give their collective opinion on the femininity of masculinity of a face. Such a process is incredibly time consuming and results in a cumbersome dataset that must be further analyzed to produce concise results.
In an effort to accelerate this research process, a multidisciplinary team of computer scientists and human anatomy experts developed a mathematical model that allows a computer to evaluate the gender of a 3D human face. The algorithm calculates how masculine and feminine a face is based up a number of different measures between major facial structures. The computer’s assessment corresponds nearly 90 percent of the time with a human’s subjective evaluation of a face’s gender.
In the immediate future, this gender-perceiving mathematical model can be used to examine the relationships between relative femininity and masculinity in a person’s health. For example, high levels of testosterone in maturing women can contribute not only to more masculine features, but also to other conditions such as polycystic ovarian syndrome. In addition, the new computer program has great potential for use in the field of cosmetic surgery where patients must be evaluated both before and after an operation.
The next step for the research team will be to apply their gender-perceiving computer to the task of analyzing the faces of children with autism. Autism is believed to be related to the levels of testosterone and other hormones that a developing child is exposed to in the womb. The hope is that the computer will be able to pick-up on facial features that would allow for the creation of an objective diagnostic measure for the purpose of placing children on the autism spectrum.
By Sarah Takushi
Sources
AsianScientist.com 
Facial Feminization Surgery
Livestrong.com
PLOS ONE
Science Daily