Abstract: In order to solve the problem of discontinuity of target tracking and large number of false alarm targets given to the same target by multi-node sonar system, a joint SVM-GAN algorithm for interrupted track association connection is proposed. Based on the correlation of acoustic features of the target track before and after the interruption, support vector machine (SVM) is used to establish the association relationship among the spatio-temporal non-overlapping tracking tracks, and the generative adversarial network (GAN) is used to continue the track set which is formed the association relationship. Meanwhile, a feedback mechanism is established to synchronically place the complete track into the training set to improve the adaptability of the algorithm to the application environment. The simulation and measured data processing results show that the proposed method can correlate the track with the acoustic characteristics of the target and track the interrupted track continuously. The correlation accuracy is more than 80%, and the number of false alarms is effectively reduced. It can be used for large range acoustic target monitoring at sea.