Extraction of nonlinear features from underwater acoustic signals based on joint prediction results

A joint prediction method based on local linear prediction is presented.Joint prediction is used to study the relation between the prediction results and the embedding parameters including reconstruction dimension and time delay.It is demonstrated that the prediction results of different signals can represent features of the signal.The joint prediction result is used in analyzing and extracting features from under-water acoustic signals.Features of underwater acoustic signals are obtained.The average slope can distinguish one class of targets from other two classes.