Abstract: In this paper,a new algorithm based on a supervised selforganizing neural network for the passive sonar target recognition was proposed.Because of the incompletement of the passive sonar sampling pattern set,this algorithm introduced a mult i-active-function struct ure and self-organizing competitive learning algorithm into the classic feed-forward neural network,and obviously improved the generalization ability of target recognition.Besides,it can efficiently reduce the learning time andavoided the local optimum.The recognition experiments of real passive sonar signals say that this new algorithm has good generalization ability and high recognition rate.