Character animation reconstruction from content based motion retrieval

We present the initial design of a motion reconstruction framework for character animation which encompasses the use of supervised and unsupervised learning techniques for the retrieval and synthesis of new realistic motion. Taking advantage of the large amounts of Motion Capture data accumulated over the years, our aim is to shorten animation production times by providing animators with more control over the specification of high-level parameters and a user-friendly way of retrieving and reusing this data, applying clustering to organize the human motion database and Neural Networks for motion generation