Located in indoor environment of wireless sensor network, the traditional RSSI (Received Signal Strength Indicator) localization algorithm has the shortcomings of inaccurate distance measurement and imprecise location because of the influence of shelter and the multipath effect. Aimed to solve the problem, a RSSI localization algorithm used RBF (radial basis function) neural network is proposed. Offline stage, the mapping relation between the RSSI value that the reference node received and its spatial coordinate is established. Online stage, the RSSI value is collected and the well trained neural network is performed to locate the node without the known orientation. The experimental results show that the proposed algorithm can effectively improve the positioning accuracy compared with the traditional RSSI localization algorithm.