Faclal Expression Recognition Based On Rough Set

Expression recognition as a typical image pattern analysis, understanding and classification of computing problems, it provides a range of specific issues of good for pattern recognition, computer vision, image processing, artificial intelligence, human-computer interaction, neural computation and psychology and other disciplines. The in-depth study and final settlement of facial expression recognition contribute greatly to the maturity and development of these disciplines. For example, as a pattern recognition problem, it is considered one of the most challenging issues:the difference between different types of models is very delicate, which is based on noise of the data collection process, the accuracy of imaging equipment, changes in external conditions and data defect and other reasons. At the same time facial expression recognition is different from other types of pattern recognition, its research is not confined to the recognition phase, analysis of human mental state and mental activities to achieve the purpose of human communication Expression recognition is more important.The smart man-machine interface is one of the core research areas, facial expression recognition is essentially to give the computer the ability to know the external behavior of people, this is one of the problems that the intelligent human-machine interface need to be solved, and that is the important performance of machine intelligence also. Facial expression recognition problem can greatly improve the final resolution of the current rigid, inflexible interactive environment for the computer to understand human thought, which to some extent, change people's way of life. There is tremendous prospects for commercial applications expression recognition, such as video conferencing, distance education, medical system and the application of computer games.The reseach is meanly based on the process for facial expression recognition feature selection method In this paper. Based on rough set theory, rough set attribute reduction algorithm, and to attribute reduction algorithm is applied to feature selection process; the same time the method of expression recognition research is contacted with artificial neural network theory in my paper. The main results obtained are as follows:1.based on the classical theory of rough set attribute reduction feature selection methods. Introduct the classical rough set theory and the rough set attribute reduction method is applied to the study of feature selection. At the same time, the rough set and neural network facial expression recognition method combining the features of the selection process had been raised, the purpose of eliminating redundant information to a large extent reduce the dimension of the feature to reduce the recognition the complexity of the process to obtain a better recognition effect. 2.The proposed reduction algorithm IMSH is based on the MSH attribute reduction algorithm. The efficiency of MSH Attribute Reduction Algorithm attribute dimension in the face of more instances of large-scale data sets is not good enough. In response to this shortcoming, the proposed improved heuristic function WS based on MSH in the rough set reduction algorithm, and then the rough set IMSH proposed reduction algorithm, the heuristic function is considered more of each WS decision rules for decision-making the impact of the overall classification table easier to find the global optimal solution rather than the MSH algorithm is easy to fall into local optimal solution. First, the heuristic function value in the event of the same case, MSA algorithm can not choose independence, and by analyzing the condition attribute set IMSA decision attribute set depends on the extent, can make a choice quickly; Secondly, on the property selection, MSA algorithm is not to be considered for the overall, only considered the largest property of support, lack of objectivity. IMSA algorithm from the overall consideration of all decision-making class for attribute selection, with good equity. The experiment confirmed that the new algorithm improves the original IMSH reduction algorithm for time efficiency and the ability to select properties.3.The FSPA-IMSH algorithm. Based on the IMSH heuristic information, the combination of the newly proposed approximation is based on the region FSPA reduction algorithm framework proposed FSPA-IMSH algorithm in my paper. The main idea of FSPA algorithm framework is:First calculate the core attribute set of decision table and set of core attributes as a starting point, based on positive approximation gradually excluded from the discourse of some examples of objects, reducing the domain space, choose to inspire type of information function value of new property added to the largest reduction focus to at the same time, by the above method repeated iterations until the reduction set to meet the termination conditions are set at this time reduction is the result of attribute reduction. In the FSPA-IMSH algorithm, the heuristic function calculation time due to the decrease of the domain space is significantly reduced. When the number of experimental samples to reach a certain size, FSPA-IMSH time algorithms have obvious advantages, which the FSPA-IMSH framework used in the algorithm is based on positive approximation FSPA attribute reduction framework, the framework of the algorithm and the heuristic information from both the two attribute reduction aspect of enhancing the speed and quality attribute selection. At the same time, the new proposed algorithm and artificial neural networks and applied to the face recognition, and there are good results in the end of experiment.