Enhanced supervised neighborhood preserving embedding for radar target recognition

Recently, the computer vision and pattern recognition community has witnessed the rapid growth of a new kind of dimensionality reduction method, namely manifold learning. Among them, neighborhood preserving embedding (NPE) is one of the most promising techniques, which can be performed in either unsupervised or supervised mode. In this paper, a new dimensionality reduction algorithm, called enhanced supervised neighborhood preserving embedding (ESNPE), is proposed. ESNPE can enhance the local within-class relations by taking into account class label information. Moreover, neighbors are found according to a new distance metric instead of Euclidean distance, aiming for better generalization. Experimental results on radar target recognition with range profiles indicate the superior performance of the proposed method, compared with PCA, NPE and supervised NPE (SNPE).