Research On Human Facial Expressions Recognition Method

The aim of human-computer intelligent interaction (HCII) is to provide natural ways for humans to use computers as aids. To interact with the human, the computer must be equipped with human communication skills, such as the capacity of understanding, distinguishing and identifying human emotional state. Facial expression includes rich information about human emotion。It is an important way of understanding emotions. In recent years, the reasons for renewed interest in facial expression recognition are multiple, but mainly due to people have more interests about human-computer intelligent interaction. In recent years, interest in facial expression recognition is renewed, due mostly to the increasing interest in human-computer intelligent interaction.This paper has analyzed and summarized some related research work on facial expression in psychology field and computer field. The previous research work was mostly based on the static human facial images and localized Facial Action Coding System. Here, this paper proposed an improved method of dynamic facial expression recognition based on Gaussian of Mixture Hidden Markov Models, by which previous system's deficiencies can be overcame, and the real facial expression movement features and emotion mentality can be reflected more truly. The main contributions are as follows:1.A method describing human facial motion features based on phase form is proposed in this paper. We get some special facial expression regions, in which the motion features are extracted and described as phase form and then constituted to time-based eigen-sequences. Experimental results show that this method simplified the process of calculating eigen-sequences and reduced processing time.2.This paper described facial expression time-based eigen-squences based on 1st-order left-right Gaussian of Mixture Hidden Markov Models. Traditional Hidden Markov Model needs vector quantization, which inevitably will introduce vector quantization error. So we used Gaussian of Mixture Hidden Markov Models to describe facial expression sequences, which describes probability distribution based on the vector space directly. And we assumed the facial expression time-based eigen-squences satisfied Gaussian of Mixture model, which reduced classification error.