Facial expression synthesis using radial basis function networks
Abstract
This chapter presents a technique that synthesizes 2-D gray-scale facial expressions using Radial Basis Function (RBF) neural networks. There are three types of networks that we have constructed to synthesize facial expressions. Facial expressions play an important role in non-verbal communication. Facial Expression Synthesis (FES) techniques try to simulate people’s expressions artificially using computers. FES can be used to gain more insight into the contribution of visual information of facial movement in relationship to speech perception permitting a better controlled and more systematic analysis of the auditory perceptual process. The backpropagation (BP) training algorithm uses the gradient descent method to seek out the minima of the error function in the weight space. After choosing the initial weights of the network randomly, the BP training algorithm computes the necessary error corrections and updates its weights accordingly. The 2-D digital image warping is a branch of image processing that deals with the geometric transformation of digital images.
Authors: Irwin King, X.Q. Li
Published in: Routledge eBooks (2022)