Facial expression recognition based on texture and shape
Wenchao Zheng, Cuicui Liu · 2016
Facial expression recognition has become key challenge in the field of anthropomorphic human-computer interaction. In this paper, an approach is presented for facial expression recognition through the shape of facial feature points and the texture information of specific areas, based on Active Appearance Model (AAM). First, find out that the shape and texture parameters can express more personalized information of each expression by analysing the physical significance of main parameters in the AAM. And then use these two features to classify expressions which rely on machine learning classification algorithm. Next, weigh the relationship between identification rate and retaining information in the training process of AAM, and understand that noise is also introduced into the classification along with the increase of retaining information. So the dimension of feature vector should be adjusted to get high identification rate in a lower dimension. Experiments show that feature extraction suitable for facial expression can get higher identification rate and is also robust, compared with the previous expression recognition directly based on the appearance parameters of AAM. Meanwhile, feature dimension can be effectively reduced under the condition of no loss of identification rate.