Synthetic multichannel time series for simulating underwater acoustic noise

The paper discusses generation of synthetic multichannel digital time series based on measured directional and spectral characteristics (second-order statistics) of underwater acoustic noise. The data are intended for analyzing the performance of array signal processors. The noise is modelled as Gaussian, stationary and homogeneous. Second-order statistics are selected as representative of noise due to distant shipping, wind action and system self-noise. A multivariate autoregressive realization of the stochastic processes is obtained via the Levinson-Robinson solution of the normal Toeplitz equations. Implementation of the time series generator on an FPS AP 120B array processor is described.