On-line registration of new users for the face recognition system using learning subspace method

摘要

In this paper, we propose an on-line new users registration method for the face recognition system using Learning Subspace Method (LSM). It reduces the user's stress to register the new user on-line. However, it is hard to recognize if there is only few data available for learning process. To solve this problem, we introduce a new registration method: calibrate the variance of data for learning by previously registered users average of variance first, then rotate the sub-matrix to separate each subspace effectively by using LSM. By using this algorithm, we decrease the recognition error of the users who registered by only few learning data.