Research Of Face Recognition Based On Incremental Learning Algorithms Of Component Analysis

Face recognition is an active research area in recent years.It has been adopted as an identity validation technology in many applications, such as information security,entrance guard controlling systems.At the same time,face recognition is a classic high dimensional data and small sample size problem.The tasks of face recognition raise strict requirements on the performance of algorithms.At present,there are bottle-neck problems in feature extraction method of face recognition, such as low accuracy,high complexity,bad robustness,worse real-time, low recognition rate.This thesis proposes Incremental learning algorithm based on component analysis to solve the above problems.We made thorough sruvey and study on discussing this problem.Principal Component Analysis(PCA)has proven to be an efficient method in pattern recognition.Recently,PCA has been extensively employed for face recognition algorithms.Howerve,PCA-based face recognition systems face an important bottle-neck problem:the effectiveness of incremental learning algorithms of component analysis. This thesis proposes a new incremental learning algorithm based on component analysis for Batch-Incremental data to solve the above problems.We apply the space projection transformation based on the original PCA,and then compute the unitary PCA in a lower transformed space.So the computation complexity can be reduced.And then we put forward the kernel form of the incremental method.We compare the IPCA algorithm with the traditional PCA,the IKPCA with the KPCA algorithm,respectively.The experimental results on JDL-A face database demonstrate the effectiveness of the proposed algorithm.Independent Component Analysis(ICA)has been proved to be a significant method in pattern recognition.It also can be used in human face recognition algorithm to estimate the independent component features of human face images.In this thesis,an incremental learning algorithm based on ICA is introduced to fit the application domain that has a high requirement for recognition rate of the system at the cost of algorithmic complexity.This algorithm calculates the principal components of a sequence of image vectors incrementally without calculating the covariance matrix.At the same time,the principal components are transformed to the independent components that maximized the non-Gaussian of the source.This algorithm is fulfilled by merging the two algorithms,IPCA and ICA,to obtain the most efficient and independent components that describe a whole set of human face database in a real-time fashion.Simulation results on ORL and UMIST databases show that the algorithm mentioned in this thesis has better real time and higher recognition rate compared to Batch PCA_ICA,PCA and Fisherface. KEY WORDS:face recognition,kernel method,incremental learning...