Kernel Methods For Support Vector Machine In The Application Of Face Recognition Research

The technology of face recognition is the use of the unique feature of differentindividuals in the face or facial expression, it is the technique that uniquely identifiesthe unknown identity,and it is a kind of computer security technology.The technologyof face recognition has the advantages of non contact, convenient and quick, highprecision, strong adaptability and good security, the application areas graduallydevelop from specialization to the broad based, face recognition technology will havevery broad application prospects.Support vector machine is put forward by Vapnik eton the basis of the statistical learning theory and structural risk minimizationprinciple,it is a kind of new machine learning method, and the method can solve smallsample, nonlinear, high dimension, and local minimum problems.Support vectormachine has been widely used in intrusion detection technology, biology informationtechnology and the many domains such as text classification. The support vectormachine classification is applied to face recognition and image processing and otherfields, have very important sense.Kernel method is an effective way of solving support vector machine nonlinearpattern classification problems, the selection and the structure of kernel function is thekey to kernel methods. The purpose of this thesis is to study the modified Gaussfunction of the effective application in face recognition.The main innovation of this paper is as follows:(1)According to the AdaBoost algorithm in face recognition, recognition time istoo long, the recognition rate is not very ideal, so we put forward SVM classificationalgorithm. The experiments verifiy that used the SVM algorithm for face recognition,can greatly shorten time and improve the recognition rate.(2)SVM with Gauss kernel function in face recognition has not very ideal effect,this paper presents a modified Gauss function, through chooseing the appropriate parameters to improve the learning and classification ability of modified Gauss kernel,and theoretically proved the legitimacy of new nuclear. The modified Gauss kernelfunction is applied to face recognition, and make contrast with Gauss kernel function,Theexperiments verify that the modified Gauss kernel function can make up for theinadequacy of Gauss kernel function, and can bring its advantages into full play.(3) In the ORL face database,the images of one person only contain differentexpression and different profile, and not contain the same person with the differenthairstyle images and the twin images.In order to show that kernel methods for supportvector machine is widely used in application and practicality, this paper establishes adifferent hairstyle face database and twins face database, and using the experimentsverify that kernel methods for support vector machine is practical and effective.