Facial expression recognition based on combined HMM

Abstract

In this paper, a new recognition method called the combined HMM is proposed. The method combines HMM and Mixture-HMM through KNN. Mixture HMMs are trained on any two classes of training samples with traditional HMMs. The recognition procedure of combined HMM is to make a final class judgement by KNN while the HMM's basic decision is different from the Mixture-HMM confirmatory sentence. At the same time, some improvements for the KNN decision criterion and the HMM interactive rule are made to enhance classification ability. Experiments show that the performance of this combined method exceeds that of HMM.