Abstract
This study focuses the expression recognition of facial images using higher-order local autocorrelation features and discriminant analysis. Higher-order local autocorrelation features are the higher-order extension of the autocorrelation function which is shift-invariant, and the range of the displacements is restricted within a 3×3 local region, the center of which is the reference point. Discriminant analysis linearly maps the primitive leaning data classified into some classes into new discriminant space, which maximizes the inner-class covariance while minimizing the between-class covariance. In this experiment, we photographs facial images of some test subjects with live basic expressions and calculates the recognition rate of expression using higher-order local autocorrelation features and discriminant analysis. We also consider the application to facial animation using the locus of expression changes in discriminant space.
