Analysis of facial motion patterns during speech using a matrix factorization algorithm

摘要

This paper presents an analysis of facial motion during speech to identify linearly independentkinematic regions. The data consists of three-dimensional displacement records of a set of markerslocated on a subject's face while producing speech. A QR factorization with column pivotingalgorithm selects a subset of markers with independent motion patterns. The subset is used as a basisto fit the motion of the other facial markers, which determines facial regions of influence of each ofthe linearly independent markers. Those regions constitute kinematic "eigenregions" whosecombined motion produces the total motion of the face. Facial animations may be generated bydriving the independent markers with collected displacement records.