Abstract

We present a data-driven animation tool that uses foot pressure sensors to control character carrying motions. During an offline training phase, we captured a high dimensional data set of synchronized multimodal motion capture consisting of sole pressure sensors and full body motion capture. Our subjects carried a variety of weighted objects. Our approach provides highly accurate recognition of both the type of carrying motion and weight of the object from only the sole pressure signal. We present a technique for synthesizing carrying motions on characters by overlaying our inverse kinematic algorithm on skeletal motion. We compare the generated animation to ground truth animation created directly from the motion capture data.