Human motion primitive discovery and recognition

We present a novel framework for the automatic discovery and recognition of human motion primitives from motion capture data. Human motion primitives are discovered by optimizing the 'motion flux', a quantity which depends on the motion of a group of skeletal joints. Models of each primitive category are computed via non-parametric Bayes methods and recognition is performed based on their geometric properties. A normalization of the primitives is proposed in order to make them invariant with respect to anatomical variations and data sampling rate. Using our framework we build a publicly available dataset of human motion primitives based on motion capture sequences taken from well-known datasets. We expect that our framework, by providing an objective way for discovering and categorizing human motion, will be a useful tool in numerous research fields related to Robotics including human inspired motion generation, learning by demonstration, and intuitive human-robot interaction.