Modeling of shallow water sea ambient noise using artificial neural network

In underwater signal processing the most important factor in quantifying the signatures of the radiating object is to decipher the signals which are prevalent in the ambient noise. Ambient noise is a complex and important phenomenon which greatly affects the listening capacity of instruments such as sonar in underwater environment. The ambient noise in sea is the overall combination of wind speed, wave speed, wave height, barometric pressure, dew point, temperature, marine life, shipping traffic and seismic activities. The work presented in this paper focuses only on three of the above mentioned parameters, namely, the wind speed, the barometric pressure and the temperature; which affect and play an important role in the overall spectrum of ambient noise in shallow water. In order to analyze the same data gathered in Ormara harbor over the past ten years was studied. Variation of the ambient noise in shallow water is investigated with respect to the above mentioned parameters. Finally a model for the ambient noise is proposed which is trained using an artificial neural network. High prediction accuracies of around 96% are obtained in different ISO standard octave bands.