Group Class Residual ℓ1-Minimization on Random Projection Sparse Representation Classifier for Face Recognition

Sparse Representation-based Classification (SRC) has been seen to be a reliable Face Recognition technique. The ℓ1 Bayesian based on the Lasso algorithm proven most effective in class identification and computation complexity. In this paper, we revisit classification then recommend group-based classification. proposed modified algorithm, which is called as Group Class Residual (GCR-SRC), extends coherency of test sample whole training samples identified rather than only nearest one samples. Our method between To reduce dimension samples, choose random projection for feature extraction. This selected computational cost without increasing algorithm’s From simulation result, reduction factor (ρ) 64 can achieve maximum recognition rate about 10% higher SRC original using downscaling method. method’s feasibility effectiveness are tested four popular face databases, namely AT&T, Yale B, Georgia Tech, AR Dataset. GCR-SRC GCR-RP-SRC achieved up 4% more accurate with class-specific residuals. experiment results show that technology group-class-based not reduces data but also increases accuracy, indicating it feasible recognition.