Classification on Underwater Acoustic Propagation Model U sing Convolutional Neural Network

Compared with electromagnetic wave, acoustic wave has great advantages in underwater sensing, which attracts great attention on how it propagates in underwater environment. A variety of types of underwater acoustic propagation models (APMs) have been developed, where the closest to actual underwater acoustic propagation should be selected as in most sensing cases, accurate APM are needed, like matched field processing, geoacoustic inversion, etc. However, selecting by experience is not reliable enough and matching the measured field with the modeled one is time-consuming. In this paper, a convolutional neural network (CNN) classifier is proposed to select appropriate underwater acoustic propagation model (APM) according to sound field collected by a horizontal line array (HLA). The simulated data are generated by five different APMs and the classification accuracy on testing set reaches 95.3%. The correlation and difference between the performance of this CNN-based classifier and the correlation coefficient (CC) of the classified data are discussed, which demonstrate that the CNN-based classifier outperforms the correlation-based classifier. We also validate the performance of the CNN classifier by the data measured in shallow water environment.