Global linear regression coefficient classifier for recognition

In this paper, a novel classifier based on linear regression classification (LRC), called global linear regression coefficient (GLRC) classifier, is proposed for recognition. LRC classifier uses the test sample and the class subspace to calculate the distance which will be used for classification. GLRC classifier uses the test sample vector and whole train space (all the class subspaces) to calculate the global linear regression coefficient. Then GLRC computes the signed square sum of the linear regression coefficients belonging to the same class, and the result will be used for classification. A large number of experiments on Yale face database and AR face database are used to evaluate the proposed algorithm. The experimental results demonstrate that the proposed method achieves better recognition rate than LRC classifier, sparse representation based classification (SRC) classifier, Collaborative representation based classification (CRC) classifier and two phase test sample sparse representation (TPTSSR) classifier and so on.