Verification of Expressiveness of Procedural Parameters for Generating Emotional Motions

Body movements are crucial for emotion expression of a virtual agent. However, the perception of expressiveness of an animation has always been a subjective matter. Research in psychology has asked professional actors to perform emotional motions to study how body gestures deliver emotions [2]. In this work, we aim to design an animation system that can generate human body motions in a more systematic manner and then study the expressiveness of the generated animation as a preceding step for the study of the relation between motion and emotion. We propose to stratify the variables relevant to describing motion and emotion into four layers (from upper to lower): emotional, style, motion, and procedural layers with their own respective sets of parameters. In the emotional layer, emotions can be modeled with either the basic emotions approach [1] or the dimensional approach. The style layer serves as an intermediate layer for describing expressiveness of an animation. In the motion layer, the parameters specific to a target type of motion are defined. In the procedural layer, generic animation procedures are used to generate parameterized motions. In this paper, we focus on designing generic animation procedures for the generation of human body motions and studying how the parameters in these procedures are mapped onto the parameters in the upper layers through the example of walking motion. In procedural animation, a motion is specified by defining appropriate keyframes and interpolations. The interpolations are performed on the procedural parameters such as joint angles or points in the 3D space. If the points in the Cartesian space are specified as a curve, the curve is re-parameterized by arc length. Given these procedural parameters, the motion parameters for interpolation are defined as the locations of the control points that are used to specify a spline curve (such as a Bezier curve) in the parameter-time space. By designing appropriate mapping between the motion parameters and the style parameters, we hope to produce expressive walking motions. We have adapted the style attributes defined in [2] for our style parameters which include smooth-jerky, stiff-loose, slow-fast, soft-hard, and expanded-contracted. For example, in our implementation, we assume that each joint is equipped with a virtual