The2D-FrFT Combined With Fuzzy Fusion Classification Algorithm In Human Emotional State Recognition

In the21st century, with the rapid development of computer science, internet, communication technology and other related fields, human emotional recognition have become a hotspot. It is widely used in safe driving, distance education, digital family, auxiliary medical and game entertainment system. The broad application prospect make the researches on emotional recognition have theory value as well as practical significance.Human emotional recognition system mainly includes several parts:image detection and predisposal, feature extraction, dimension reduction and classification. The purpose of image detection and predisposal is to get the human emotional image, by the predisposal, make it more conducive to the recognition system. Then the feature extraction is to get the emotional feature by some algorithm or transform. While dimension reduction is on the conditions that preserve feature information, decrease the complexity of the recognition algorithm. The last step which is important called classification, by the analysis of feature information and classify algorithm, give the recognition result. The papers do the research on feature extraction, dimension reduction and classification. The specific work is as follows:We perform2D-FRFT on the feature extraction of emotional image. Firstly we introduce the definition and theoretical of FRFT. According to normalization dimension we draw the discrete FRFT and then derivate the2D-FRFT and its discrete form. Then we perform2D-FRFT on the feature detection, make research on the amplitude and phase information.We utilize PCA and FLDA method to decrease the high dimensions of the emotional information after feature detection. PCA have the advantages as:energy maintain, de-correlation, energy redistribution, so maintain the important information which made it easy to processing. While FLDA algorithm made the between-class distance maximize and the in-class distance minimize, which may further decrease the high dimensions.The last parts on classification, we propose a fuzzy fusion algorithm based on the2D-FRFT. Firstly, we utilize three single classification to classify the emotional information, get a confidence matrix. Then perform fuzzy integral on the confidence value and fuzzy measurement. Simulation results show that, the fuzzy fusion algorithm show good performance.