4D (3D Dynamic) statistical models of conversational expressions and the synthesis of highly-realistic 4D facial expression sequences