Linear regression classification (LRC) has attracted a great amount of attention owning to its promising performance in face recognition. However, its performance will dramatically decline in the scenario of limited… Click to show full abstract
Linear regression classification (LRC) has attracted a great amount of attention owning to its promising performance in face recognition. However, its performance will dramatically decline in the scenario of limited training samples per class, particularly when only single training sample is available for a specific person. In this paper, a novel LRC based method is proposed to solve this problem. Specifically, we first perform LRC on the training set to obtain a kind of residual for each class. Next, a reverse representation residual is derived for each training sample of a class by exploiting the linear combination of the training samples of its nearest classes and the test sample. Then, we combine the reverse representation residuals of the class by an adaptive weighted-average approach to produce the other kind of residual. Finally, two kinds of residuals are fused to classify the test sample. Experimental results on the ORL, FERET, Libor94 and CMU-PIE face databases demonstrate that the proposed method obtains a higher recognition rate than some state-of-the-art face recognition methods.
               
Click one of the above tabs to view related content.