Local linear regression classifier for image recognition

In the past several decades, much work has been done to design classifiers. Inspired by the locality idea of manifold learning, a local linear regression classifier (LLR classifier) is given in this paper. The proposed classifier consists of three steps. The first step is to search k nearest neighbors of a test sample from each special class, respectively. The second step is to reconstruct the test sample based on the k nearest neighbors from each special class, respectively. The third step is to classify the test sample according to the minimum reconstruct error. The proposed local linear regression classifier is evaluated on the CENPAMI handwritten number database, the ORL face image database and the ORL face image database. The experimental results demonstrate that an LLR classifier is effective in classification, leading to promising image recognition performance.