There is large error caused by the burst noise of the received signal strength (RSS) in the indoor fingerprint positioning. To address this issue, a subspace matching method combined with the algorithm of clustering by fast search and find of density peaks (DPC) is proposed. In this way, the large positioning error could be effectively avoided. Specifically, the coverage vector of the access points in the online RSS is used to select a subset of the reference points and the subspaces of the radio map. An improved weighted K-nearest neighbors ( WKNN) approach is applied to achieve estimation. Then, the DPC algorithm is used to select S estimated positions with the largest decision values to determine the position of the target. The simple algorithm needs no learning process in the offline stage, which is especially suitable for large indoor areas with a lot of access points. Compared with the WKNN algorithm, experimental results show that the proposed method improves the positioning accuracy by about 25% . The large positioning error of more than 4 m is eliminated when the reference points is 1. 8 m×1. 8 m. The overall positioning performance is improved effectively.