"Warm Bodies": A Post-Processing Technique for Animating Dynamic Blood Flow on Photos and Avatars

TL;DR

This paper introduces a data-driven method to animate subtle blood flow patterns in avatars and photos, enhancing their lifelikeness and emotional expressiveness by simulating physiological signals.

Contribution

It presents a novel physiological model for animating blood flow, improving avatar realism and emotional perception through subtle physiological animations.

Findings

Participants preferred faces with blood flow animations as more anthropomorphic.

Manipulating heart rate frequency alters perceived arousal levels.

Blood flow animations increase perceived lifelikeness of avatars.

Abstract

What breathes life into an embodied agent or avatar? While body motions such as facial expressions, speech and gestures have been well studied, relatively little attention has been applied to subtle changes due to underlying physiology. We argue that subtle pulse signals are important for creating more lifelike and less disconcerting avatars. We propose a method for animating blood flow patterns, based on a data-driven physiological model that can be used to directly augment the appearance of synthetic avatars and photo-realistic faces. While the changes are difficult for participants to "see", they significantly more frequently select faces with blood flow as more anthropomorphic and animated than faces without blood flow. Furthermore, by manipulating the frequency of the heart rate in the underlying signal we can change the perceived arousal of the character.