Cns2002- Submission Extended Abstract Nonlinear Dynamics and Motion in Human Face Image Space

The main contribution of this research is two-fold: (a) we provide a rigorous substitute for the intuitive regularity observed in visual perception, by appealing to information geometry; and, (b) we provide a set of algorithms for estimating measures of curvilinear invariants of data space through nonparametric statistical estimation of natural dynamics within the set of images. We relate the nonlinearity of the image space of human faces to temporal dynamics in facial movements and other optically detectable variations