Contrast one of the Key Issues in Determining Both Avatar Facial Gender and Age
In the human mind, the perceived difference between genders and ages is what matters. How the eyes process what they see and turn an androgynous face into one gender or another, a young face into an aged one, depends to a surprisingly great deal on subtle cues.
Take the two faces below, for instance:
A quick glance tells you that the one on the left is female, the one on the right male. Something in the way each is shaped, gives it away. Except it doesn't. They're both two pictures of the exact same face, both originating from the same original digital photograph. The difference lays entirely in how they have been treated.
No facial features were moved or airbrushed out here. Instead, the striking difference was achieved through altering the contrast levels. That's it. Alterations in contrast between select areas of the face was the only thing that changed. Specifically female faces have greater contrast between eyes, lips, and surrounding skin than do male faces. Alter the contrast levels, and you alter the perceived gender.
The work of Psychology Professor Richard Russell of Gettysburg College unearthed this fascinating fact back in 2009. His work published in the journal Perception, unearthed that the primary way our brains differentiate between genders is a matter of contrast perception. He went on to demonstrate this by, starting with volunteers with androgynous faces, altering the contrast levels around key areas and showing the resulting images to a second group of study participants.
By measuring photographs of men and women, he found that female faces have
greater contrast between eyes, lips, and surrounding skin than do male faces.
This difference in facial contrast was also found to influence our perception
of the gender of a face.
His work was originally intended to be used by the cosmetics trade, showing how cosmetics can be used to amplify the differences correctly and make a male face appear more feminine or a female face more masculine by the correct application of cosmetics.
But, here we can go one better. We don't work with cosmetics; this is VR. The avatar of the user does not have to start with their physical face, although a fair few systems now, will wrap a photograph of the user's face and head around a blank avatar head. At other times, a new face is created from scratch, especially for that person.
In either case, by taking advantage of this work on contrast, especially the areas where contrast is most heightened, we can control the gender an avatar face is perceived as. This can both help those with androgynous faces who have applied their face to an avatar wraparound, and wish to be able to apply subtle clues as to their preferred gender on it; or it can be used equally well to create a face that through subtle cues only, appears to be the desired gender. This includes the possibility of reusing the same basic face for multiple avatars, and simply altering the contrast as necessary to create the avatars for related individuals.
As stated above, the initial research in the psychology of faces by altering the contrast levels of various aspects is over four years old at time of writing. So it should come as no surprise that the work has been pushed even further. Specifically, a 2013 paper has shown how it is not just gender that altering the contrast levels affects. Gender is altered by altering contrast levels around specific features. If you alter other parts of the face, you change the perceived age of the face.
This has evn more useful possibilities than the first finding: We now have the possibility of a user's avatar aging with them, taking on appearance of being older as they use it year on year, without them having to consciously remember to change the image that forms the foundation of their face. It alters with them in other words, like a physical face would do.
This work has also been led by the same person, psychology professor Richard Russell. Again he was using it with a focus on the application of makeup, and again it has wider implications for use in VR where we can create a face from scratch or digitally alter a source image from a user to our heart's content.
The discovery of this cue to facial age perception may partly explain why cosmetics are worn the way they are, and it lends more evidence to the idea that makeup use reflects our biological as well as our cultural heritage, according to Russell.
Unlike with wrinkles, none of us are consciously aware that we're using this cue, even though it stares us in the face every day, said Russell.
In one study, Russell and his team measured images of 289 faces ranging in age from 20 to 70 years old, and found that through the ageing process, the colour of the lips, eyes and eyebrows change, while the skin becomes darker. This results in less contrast between the features and the surrounding skin leaving older faces to have less contrast than younger faces.
The image above illustrates this very well. As before, these two faces have come from the same original digital photograph. Yet the one on the right looks much older than the one on the left. As before, the difference was achieved entirely through altering the contrast. The facial contrast has been increased in the left image and decreased in the right image.
Since the altering of contrast is such a simple process, and with modern facial recognition techniques being widely available in commercial software packages and under license in raw form, it should be a relatively simple process to create an avatar facial modification process using these details.
Taking things a step further, these details embedded in the virtual environments access software, would make it quite possible for subtle changes to be applied to the avatar of users over time. Say by applying a very slight contrast adjustment every six months or so. Over a period of years, a user's avatar would age apace with them, greatly increasing the feeling for the user that their avatar truly both reflects who they are and is a part of them.
By being aware of which elements contribute towards gender identity as opposed to age itself, it is equally relatively easy to not apply contrast changes to accidentally alter the perceived gender whilst altering the perceived age. That is where the facial recognition aspect would come in; picking out the pieces that are not to be altered lest the user find they have apparently swapped genders without warning as time moves on.