Research On The Method Of Face Recognition

As a biometric technology, Face Recognition Technology has numerous applications. It is emerging as an active research area in the field of pattern recognition and artificial intelligence. Face Recognition is important for surveillance and security, telecommunication, digital libraries, video meeting, and human-computer intelligent interaction.In this paper, a Face Recognition method base on principle component analysis (PCA) algorithm was investigated. We study the rationale and the whole process of PCA when it was used in Face Recognition, and we analyse its excellence and disadvantage, then we introduce a method called two-dimensional principal component analysis (2DPCA), it reserves the frame information of face feature, so the accurate recognition was improved. Because of the influence of some factors, such as brow, illumination and the change of face feature along with people's age increase, Face Recognition is not a problem of simple linear classification, so we introduce a method called kernel principal component analysis (KPCA). This method makes the classification near to linear classification by the transformation of feature space, so it improves face recognition rate. It was proved by large numbers of experiment. In practice, sometime face image was introduced in real time, it needs to chang image feature each time. So we introduce a method called candid covariance-free incremental principal component analysis (CCIPCA). This method computes the principal componernts of a sequence of image vectors incrementally without estimating the covariance matrix.The low-frequency subband of wavelet transformation contains the primary information of the original image, and the brow of face incarnates by eyes and mouth. Both wavelet transformation and PCA can reduce dimension, but they all have localization when we use each of them only, so in this paper we make low-frequency subband combine with horizontal-frequency subband as training samples then use PCA, 2DPCA, KPCA, CCIPCA methods which have been mentioned. For wavelet transformation can reduce dimension and interfere information, so improve face recognition rate and speed.Face recognition is usually a small sample problem of pattern recognition, but feature extraction and classification need enough samples for training can give good result. In this paper we use the given images creating new sample, increasing training sample's number, then satisfy the need of classify algorithm. This method improves the small sample problem in a certain degree. In this paper we use all the methods which have been mentioned to experiment, analyse the result and compare with their capability. We complete investigate the Face Recognition method base on principle component analysis (PCA) algorithm.