Representation and Recognition of Complex Human Motion

Jesse Hoey and James J. Little

Abstract

Automatically recognizing complex human motions such as facial expressions is a difficult task. It is beneficial to have a simple and yet general representation on which to perform classifications. We present a holistic model-free representation of motion which uses the complete basis of orthogonal Zernike polynomials. This representation is simple, yet general, and can be used for describing many types of motion at many scales. Starting from image sequences, locally smooth image velocities are derived using a robust estimation procedure, from which are computed compact representations of the flow using the Zernike basis. Continuous density hidden Markov models are trained using the temporal sequences of vectors thus obtained, and are used in subsequent classification tasks. We present results of our method applied to image sequences of facial expressions both with and without significant rigid head motion and to sequences of lip motion from a known database. The results demonstrate its effectiveness at different classification tasks.