Emotions in games: More sophisticated than this.

Research by the University of Abertay into facial expressions could hold the key to making computer game characters with more realistic facial animations and emotions.

Robin Sloan, a PhD student and lecturer based in the University’s Institute of Arts, Media and Computer Games, has devised a set of rules that could help portray a more convincing facial animations and emotions. These rules will no doubt be very useful to Train2Game students, especially the Artist & Animators!

(Game Designers and Game Developers: you can read on, or watch the impressive F1 2010 Developer Diary that was pointed out yesterday)

The study involved a series of experiments examining how the upper and lower regions of the face move during expressions including happiness, surprise and anger.

The aim was to make every stage of the choreography as believable as possible. Actors were used in order to study realistic expressions induced by genuine emotions.

It was found that for sadness to look real, it needs to lead from the upper face with, the furrowing of the brow and lowering of the eyes should occurring before the mouth corners turn downward. If this expression unfolds the other way round, the study found it looked childlike or faked. (I’ve found the latter often occurs during Dragon Age: Origins, though the rest of the game is excellent!)

Similarly, for anger, initiating the expression with the upper face works best in practice with the lower face following thereafter – rather than gritting one’s teeth alone.

Choreography can also affect how clear the emotions are when observed by audiences, such as the gamer. For instance, disgust animations may look fairly authentic when the upper face leads, but the lowering of the brow can result in the expression being confusable with anger. In this case, leading with the lower face creates a more distinct disgust expression.

The team also studied emotional expression transitions, for example from happiness into sadness, or sadness into anger. Robin Sloan explains the findings:

“What we found in this second stage of the study was, for example with surprise into happiness, if the upper face moved before the lower face, this could result in an insincere happy expression which could be viewed as an exaggeration or, indeed, fake. This could be useful if animators deliberately wanted to create a fake smile, but would otherwise be unhelpful.

“On the other hand, when the lower face led the movement in this transition, the overall animation appeared much more believable. Likewise, for happiness into sadness, upper face leading seemed clear and credible, whereas leading with the lower face seemed childish or sarcastic, as if displaying an interpretation of sadness rather than genuinely portraying the emotion.”

He continued: “While much is known about the appearance and perception of emotional facial expressions, researchers and professionals still struggle to create perceptually believable animated characters. For example, films such as Polar Express and Beowulf are ‘performance-captured’ where the performance of human actors is transferred onto computer animated characters.

“However, the aesthetic results of this technique have not been fully embraced by the public, as it appears that audiences view the characters as fake and unrealistic. Indeed, we are often more likely to believe in characters from more traditional animation films such as Toy Story or Shrek – animations which are carefully crafted by teams of animators.

“While the computer animation research community is quite rightly interested in the technical possibilities of performance capture, we wanted to highlight the fact that traditional animation can still play an important role in research, and to show that an artistic approach to animation can yield tangible research findings. We feel that our research could, for instance, have implications for the development of believable computer game characters, as an understanding of what makes for believable facial expression animation can boost their credibility.”

Mr Sloan hopes that the results could be useful for Games Designers, Game Developers and Games Animators – like Train2Game students – seeking to create more believable and, more interactive characters.

The research was published in the Journal of Computer Animation and Virtual Worlds.

So, Train2Game universe,  what do you think of the study? How much have you thought about how animation works in your games? And what research do you do before animating characters?

As usual, leave your comments here or on the Train2Game forum.