Abstract
The extraction of inexpressive facial information has been carried out by applying RGB convolution (i.e., the convolution of each color component, Red Green and Blue) to the Saito Fourier-Wavelet transforms. Pictures of a child's two different facial expressions, normal and smiling, were transformed separately for each color component. The data from this transformation was analyzed, to find differences between normal faces and smiling faces, which were commonly observed in the convolution of the three colors. RGB convolution could show slight differences in facial information that the monochrome data could not reveal This method has considerable potential for extracting slight facial information of autistic children.
