The Study Of Facial Expression Recognition Based On The Improved PAD Emotion Model

With the development of science and technology, computer plays an increasingly important role in people’s work and life. If the computer has the ability to understand and express emotion, and adapt to the environment independently, it will fundamentally change the human-machine relationships, enhance harmonious, cordial and accuracy of human-computer interaction, and possibly achieve the emotional interaction between human and computer. This will be a great help for improving people’s work efficiency and work space, and enables the computer to serve mankind better. Facial expressions can convey very rich and delicate facial information, and therefore characterized by their unique facial feature to distinguish the user identity and access to relevant information. Facial expression recognition technology has been the concern of academics and become a hotspot. This dissertation focuses on the fine classification of expression using of the improved PAD-based emotion model. The main work of the paper is as follows:First, how to improve the PAD-based emotion model is studied. This paper analyzes the research status of the emotion model. PAD theory of emotion has been widely used in the field of personality psychology, emotional psychology and social psychology and other fields as well as product satisfaction, marketing and so on. The projections of emotion in the three-dimensional PAD emotion space are discrete points. According to our life experience, we know that our feelings will not mutate, but change continuously. Therefore, we propose the improved PAD-based emotion model, which is centered on the basic emotion categories.Secondly, how to extract facial expression feature more effectively for the improved PAD-based emotion model is studied. There are many expression feature extraction methods. We select a method of combining active shape model (ASM) and gray level co-occurrence matrix (GLCM) to extract facial feature, from the existing facial feature extraction methods for the improved PAD-based emotion model. In order to extract the deformation of facial features organ, which is caused by the expression changes, this paper defines16key feature points, uses active shape model to locate these feature points, and extracts deformation features and texture features for the next fine classification.Thirdly, how to classify the facial expressions more detailed using the improved PAD-based emotion model is studied. Using the facial expression features from the previous step, depending on the size of facial features regional deformation, classify the facial expressions more detailed.Finally, we propose a non-contact health monitoring system (NCHMS). The system collects the user’s facial expressions, movements and sounds through cameras, digital cameras and other equipment, under the premise of not interfering with the user’s daily life. Then the system analyzes the collected information and determines the user’s current situation. Once the user may be at risk or is in danger, the system reminds the user immediately, and notifies the hospital, nurses and family members to take rescue measures.Our research aims to improve the PAD-based emotion model and the experimental results show that the improved PAD-based emotion model can classify the facial expressions more detailed and can be applied to the non-contact health monitoring system.