The Uncanny Valley The Effect of Removing Blendshapes from Facial Animation

Ever since its introduction in 1970 by the roboticist Dr. Mori, the uncanny valley has become an integral part of facial animation. While most of the work on the uncanny valley is focused on finding a way to surpass the uncanny valley and how certain things could be designed to increase or decrease the uncanniness of an animation, this paper investigates how removing certain blendshapes from the facial animation changes our perception about the eeriness of that animation. The goal is to identify the “necessary” blendshapes which bring about a significant change the perceived uncanniness of the animation. For achieving this, participants undergo visual tests which entails observing short animated clips expressing different emotions and recording the human-likeness and uncanniness, as perceived by the observers. The results validate some of the existing theories about the uncanny valley and also identify the necessary blendshapes for every emotion. The results show that removing the blendshapes not only affects the eeriness but also the perceived human-likeness and may change our perception of the emotion expressed in the animation