A New Sparse Representation Classifier (SRC) Based on Probability Judgement Rule

Sparse representation classifier (SRC) is a classical method for classification, which was proposed in 2009. SRC method boosted the research of sparse representation, based on SRC, many other new methods were also proposed, such as structured-sparse representation, collaborative representation optimized classifier (CROC), deformable sparse recovery and classification (DSRC) method, misalignment robust representation (MRR) method. In recent years, many classification application problems were solved by using above methods. Although SRC is effective for handling with the classification applications, it also has some disadvantages. The classification criterion of SRC is from the residuals using truncation way. If the ith residual is the smallest, then the SRC judges that the test sample belongs to the ith class. However, from the truncation way, some test samples are also misclassified by SRC. This paper presents a new probability judgement rule to sparse representation classifier (SRC), which using a probability judgement model instead of using the truncation residual directly. The extensive experimental results show that the proposed new method can achieve higher classification accuracy rates than SRC using truncation residual directly.