Synthetic signals for signal processing

Modern signal processing algorithms exploit spatial and temporal features of signals and such algorithms can be highly sensitive to structured noise. Measured data is routinely used to evaluate algorithms but this approach is often limited by the expense and impracticality of collecting data and the presence of unknown events in the data. It is often necessary to use simulated signals as an interim stage. The paper introduces a method to generate spatially correlated noise as a first step towards developing a fast and efficient scheme to generate non-stationary, spatially and temporally correlated noise. This method is tested on standard (Bartlett) and high resolution (minimum variance) beamforming algorithms to demonstrate how the performance is changed and the advantage to using correlated noise to evaluate signal processing algorithms which require more realistic test signals.