Functional body mesh representation, a simplified kinematic model, its inference and applications

The paper describes a simplified representation of a body structure and a GMM based method for inferring from the motion capture data based on a functional relationship between the points. The proposed representation can be efficiently used for marker-wise processing of the data. The parent-child and sibling relationships are inferred on a coherence of movement and constancy of distances. For creating groups representing specific body parts we propose an incremental multicriterial clustering algorithm employing Gaussian mixture models. To infer body parts hierarchy we propose utilizing a consensus method.