Hybrid classifier for marine vessel based upon propulsion type

Abstract

Finding features which can reliably classify the ships using radiated acoustic signals is an important and classical problem. It becomes more challenging when the signal has to traverse the underwater channel to reach the receiver. This is because both ship radiated acoustic signal as well as underwater channel change very rapidly with time and frequency. Moreover underwater channel is non-stationary, impulsive and has severe multipath. In this paper, a new hybrid classifier combining various features and classifying algorithms, to classify the marine vessel based upon their propulsion (diesel, gas, steam) has been suggested. Computer simulations show that suggested method works well even at low SNR (signal to noise ratio) and with severe channel.