Font Size: a A A

Face Verification Based On Hidden Markov Model

Posted on:2010-01-23Degree:MasterType:Thesis
Country:ChinaCandidate:K F ZhuFull Text:PDF
GTID:2178360275473075Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
ID authentication and recognition are most important activities in daily life.During the development of society and requirement,national public security and information security need more credible ID authentication technology.Authentication technology based on face feature is an important method to meet this need.Since Hidden Markov Model can represent correlation between face organs,it can be applied in face modeling and verification.The contributions of this paper include: HMM method is applied for face verification.The effect on verification rate of several observation vectors optimizing methods is explored.And a face verification system is designed.The principal contributions are listed as follows:1.For face verification,the size of observation vectors for every client is respectively small and the computing time needs no more consideration,therefore, advanced HMM observation vectors is important for the performance of the algorithm. An HMM observation optimizing proposal based on several feature extracting method is presented.It includes:Gabor,Wavelet features,DCT and K-L transformation. Experiment shows the verification rate promotes.2.Embedded Hidden Markov Model is adopted with the optimized observation vector to build and train the model for 2-D face images.Then face verification is actualized by using the trained E-HMM.A dynamic threshold adjustment method is presented in verification,which promotes self-adaptive ability of the model and receives good verification performance.3.System design to some extent affects the performance of the core algorithm. Good system design can make the application more effective.In this paper,a normalized face capturing method is presented,a design for locale environment of face verification is proposed,and an ID verification system based on face information is actualized.In this paper,HMM is applied into the area of face verification,the HMM's observation vectors are optimized and a dynamic threshold adjusting method is proposed.The experiment shows good verification result.An ID verification system based on face information is actualized,and the future research aims are presented.
Keywords/Search Tags:Face verification, Hidden Markov Model, Feature extracting, Gabor, Waveletl PCA, DCT
PDF Full Text Request
Related items