A frequency-domain approach to simulating ambient noise fields

To quantitatively evaluate the performance of a sonar processing algorithm, it must be tested against a variety of ambient noise fields. Although measured data is the ultimate test, simulated data provides the opportunity to examine the algorithm's performance for well-defined fields, To this end, a Monte Carlo approach was used to simulate a three-dimensional ambient noise field possessing realistic cross-spectral properties. Weighted monochromatic sources are distributed about an array of sensors to create the desired directional characteristics of the noise field. The hydrophones comprising the array can be positioned naturally in Cartesian coordinates to form any desired array structure and orientation. Noise signals are represented in the frequency domain where phase shifts due to propagation are most naturally applied. Examples of simulated cross-spectra for a variety of noise fields will be presented, and good agreement with theory is found for scenarios in which analytical solutions are possible.