A new manifold learning algorithm named adaptive neighborhood preserving discriminant projection method is proposed for the feature extraction of high-range resolution profile (HRRP)-based radar target recognition. By utilizing the objective functions of both neighborhood-preserving projection (NPP) and adaptive maximum margin criterion (AMMC), the proposed method can not only preserve the neighborhood structure of original data in the dimensionality reduced space, but also exhibit good classification performance. The proposed method is applied to the feature extraction of HRRP-based radar target recognition. Numerical experiments show that the proposed method can effectively reduce the dimensionality of HRRP and give satisfactory recognition rate.