Research On Human Facial Expression Approach Based On Statistical Feature Extraction And PSO

With the development of new information technology and multimedia, human-computer interaction (HCI) has been developed dramatically. It is not only depends on the traditional computer peripherals, such as keyboard, mouse and displayer, but more relies on the multimedia ways including images and videos, which helps the HCI more efficient and friendly. Due to the increasing computer performance and declined cost of image acquiring and processing, the computer vision system can directly run on the desktop system and even can be embedded in the micro-system. All these issues make it possible for the computer to acquire and understand human face expressions.Facial Expression is a kind of effective method to convey and communicate emotions. Expression perceiving by computer, according to the perceiving habits of human, is becoming to an important method for HCI. Human facial expression recognition is an interdiscipline of psychology, bio-informatics, image processing, computer vision, pattern recognition and machine learning, which has been paid more and more attention from its emergence in1980s.The human face images are so complicated that the interference factors, such as face position, lighting, viewpoint and different people, should be removed. Therefore, amounts of classical algorithms for face detection and localization, expression features extraction and classification have been presented, most of which have been turned out to be effective. After learning these successful methods, this paper proposed the human facial expression recognition approaches combining the statistical features extraction and optimization algorithm. These approaches take full advantages of the multi-resolution property of wavelet transform, low cost of statistical features extraction and the efficient optimization power of particles swarm optimization.This paper mainly focuses on the expression features extraction and optimization related theories. Specifically, it contains the details as follows:1. Introduce the background and meaning of the subject. And analyze the basic theories and techniques of computer vision and pattern recognition briefly. Introduce the background and the state-of-art of human facial expression recognition methods. And then analyze the parts of facial expression recognition system and the corresponding methods to realize the function the each part.2. Propose new approaches based on the research above. In general, the wavelet transform is used to reduce the data of expression images, maintaining the useful components for FER. And then the PCA and LDA methods will be used to reduce the image dimensions and extract expression features. Furthermore, the PSO algorithm and its improved form are introduced to optimize the features acquired above. Finally, take use of the nearest neighbor cluster method to classify the expression. And the experiments performed on the JAFFE database turn out that the methods proposed in this paper can get better results compared to other methods.3. Summarize the materials in this paper and identify the research area in future. The research emphasis will be transferred to the semi-supervised learning and unsupervised learning expression recognition, as well as the micro-expression, hybrid-expression recognition field.