Abstract: A passive acoustic localization algorithm based on RBF neural network was proposed.According to the principle of TDOA, four-cross array was chosen.And the fast study and non-linear approach capacity of the RBF neural network was made use of to realize rapid and precise localization.Furthermore, relative analysis has been done among RBF neural network, BP neural network, WLS algorithm, Chan algorithm, and Taylor algorithm.The performance of RBF neural network was simulated.The simulation results indicate that algorithm can improve the localization accuracy and robustness, and its performance is better than other algorithms.