Research on the nonlinear modeling and prediction method of underwater acoustic signals

Underwater acoustic signal with nonlinear dynamic characteristic has a very important research value to the prediction and filtering. The prediction and filtering for underwater acoustic signal, especially the reverberation and background noise, is the foundation of underwater target signal detection, and has important application in the non-stationary, non-Gaussian and nonlinear underwater acoustic signal processing. According to the typical linear theory of least square estimation and the Volterra series theory, the models of target signal are established respectively to process the further one-step and multi-step prediction, after comparing and analyzing the predict results, the optimal prediction parameters are selected. The prediction results show that for predicting the underwater acoustic signal, the prediction relative error of Volterra filter model based on the singular value decomposition is an order of magnitude smaller than the method of least square estimation, and the prediction results more close to the real values.