Extraction And Recognition Of Expression Feature Based On Sequence Images

Facial expression recognition is to analyze and detect the special expression state from given expression images or video frames and then to ascertain the subject's specific inborn emotion, achieving smarter and more natural inter-action between human beings and computers. The study of facial expression recognition based on video frames has found important realistic values, we make research in three steps of it. Our research work focus on the following aspect:Combining the psychological research results, in this thesis, we have set up one basic facial expression database which includes 120 video sequences at first .We present a feature extraction algorithm of optical flow by KL transform. This paper extract the feature regions of the expressions based on the facial physics-muscle model and evaluats the optical flow of the expression image sequences. The eigen-flow vectors can be calculated to constitute the eigen-sequences, and therefore, the expressions can be analyzed. Feature vector extracted from optical flow is always of high dimension. So we use KL transform to low the dimension of the optical flow based feature vector, then use the transformed vector for facial expression classification.A method based on the Hidden Markov Model(HMM) is presented that uses the optical flow feature vector as the observation vector. Left-right HMM model is used in sequences images. It gains a HMM model for each expression. A sequence features reach the HMM model. The biggest probability is the corresponding expression.