The Research Of Facial Expression Recognition Based On LBP And Deep Learning Model

So far in the field of facial expression recognition, more and more attention of scholars are focus on searching for a method which is accurate, high-speed under the multi-pose and complex environment, and it has become a hot research field. Facial expression recognition has wide application prospects in telemedicine, traffic monitoring, robot service, and other Human-Computer Interaction(HCI).Improving the rate of facial expression recognition in complex environments has important theoretical as well as practical significance.After decades of development, problems are focused on two aspects: the feature extraction algorithm and the classification algorithm. They are both play a decisive role in improving the accuracy of facial expression recognition.In this paper, in-depth research of facial expression recognition system is conducted. According to the current technique of feature extraction in facial expression recognition, we analyzed the LBP feature and proposed a method of multi-feature fusion based on LBP in order to ensure the feature of expression to be robust and simple, as the meaning time to conclude the expression feature as complete as possible; According to the main algorithm of current classification, we groundbreaking proposed that deep learning algorithm can be applied to facial expression recognition. We made an attempt and done several experiments, and achieved satisfactory results in the recognition.In this paper, we designs a method of LBP feature extraction and fusion under JAFFE and CK+, using two linear classifier SVM and k-NN and deep belief network to conducting face recognition experiment. Comparative experiment and analysis also was done after this.