A neural network for sonar target recognition

Abstract

This paper explores the area of sonar target recognition us­ ing a feedforward neural network trained with the backpropagation algorithm. Sixteen sonar waves from a variety of targets were used in the experiment Noise was introduced to each waveform at 13,10, and 3 db signal to noise ratio. The network was trained with different com­ binations of the noisy waveforms and tested with noisy data not used in the training process. 100% correct classi­ fication was obtained in all cases.