Research On Several Key Methods Of Face Recognition

Face is one of the most important biological features, presenting much important biological information, such as identity, gender, ethnic, age, expression etc. With the development of computer science, computer vision and pattern recognition based on facial images has been widely studied and performed well, which play an important role in security surveillance, artificial intelligence, human-computer interaction, and electronic commerce etc. In this paper, various issues based on face recognition were studied.Based on integral image proposed by Viola, some new rectangle features were proposed for presenting face and eye feature. Combining eye geometry feature, an automatic frontal face and eye detection system was proposed and achieved efficient results.Active shape model (ASM) is considered as one of the most useful methods for face alignment. In this paper, we proposed an improved scheme for landmark points search, making the search procedure more veracious and efficient. Skin color likelihood was introduced to build the appearance model. Experiments based the FERET face database shows that our algorithm has a higher accurate probability.A combination of Gabor, Adaboost and SVM was adopted for gender recognition. Compared with the other methods, PCA and SVM, ICA and SVM, Haar-like, Adaboost and SVM, the proposed method showed preferable results.Investigating existent algorithms for facial expression recognition, this paper proposed anew face presentation——PPBTF( pixel-pattern-based texture feature). Compared withGabor, PPBTF also presents facial texture information effectively. The proposed feature is robust to illumination and speedier in computation time. In this paper, using NKDA based on RBF kernel, we proposed a face recognition system. Compared with the other typical methods, PPBTF method achieved better recognition rates and preferable characters in real-time applications, which shown PPBTF is an effective face presentation.In the last section, application of face recognition in mobile electronic commerce was discussed. This paper thoroughly elaborated mobile e-commerce and mobile payment, especial the security problem. Face recognition performance is determined by two crucial factors, both algorithm and system design. Combining the characteristic of mobile terminal, this paper realized an identity authentication system based on face recognition technology, meanwhile analyzed the factors that were significant to authentication performance.