| Part of face recognition is the facial comparison, which is to compare with the sampledinformation of faces and find out the closest matching object. It is one of the key issues ofpattern recognition. According to the feature selection of pattern recognition, the selected partof the face information can be used in face recognition. Feature selection is an important stepin face recognition process, with a variety of methods, complicated or easy. Feature selectiondirectly influences the design a performance of the classifier. Firstly, this paper makes a briefintroduction on three classical methods of feature selection and a search technology used inthe paper which is named as the best individual feature selection. Based on the best individualfeature selection, appropriate criterion can improve the result. Therefore, this thesis appliesthe maximum information coefficient (MIC) as the validity criterion. By this, the definition,properties and algorithms of maximum information coefficient are introduced. Finally, thismethod is applied to face feature selection and recognition in order to test the validity. |
