Motion analysis on human emotions for 3D walk cycle animation

Abstract

This research studies the motion graph profiles of happy and sad walk cycle to provide references to animators in achieving realistic emotional walk animation. Data of happy and sad walks are obtained from a real person via motion capture method and they are analyzed to learn the specific attributes of the motion graphs that are associated with a particular emotion. The analysis reveals that happy walk is characterized with a double bounce at the hip joint. Sad walks exhibit similar property, but has an opposite timing and acceleration patterns. Happy emotion consistently displaying low initial acceleration before a sudden increase in speed as the leg action ends. Sad emotion, however, has smoother transition from initial high acceleration to low acceleration. The analysis also reveals that happy and sad motion characteristics are opposite to one another. The observation on a series of key poses shows the similarity in the motion characteristics to Densley’s (1999) emotional figure animation graph models. The repeated presence of certain motion patterns in the motion capture samples leads to the modelling of the motion graph specific for emotional walk cycle animation. The graph models are tested on several walk cycle animations with different walking styles. Although the resulting animations do not portray a specific emotion, the rhythm of movements that are associated to the emotion is consistent and similar to the motion capture samples. This concludes that the motion graph profiles can indeed influence the attitude of the animation without intervening the creative input from the animator.