High frequency noise modeling: Dependence on application

Abstract

It is tempting to think that one could build a practical, general purpose noise model suitable for high or low frequencies and any application. For many signal processing and system applications this may be true, and one could incorporate directionality and spatial coherence. Recent developments in the use of wind-and sea-surface noise to determine geoacoustic properties by cross correlating hydrophone signals or beams demonstrate the need for more detail. The correlation peaks that correspond to absolute distances or ranges clearly require (frequency domain) coherence in any simulation. In contrast, layer information derived from angular reflection loss measurements by spectral factorization does not require coherence since it originates from an angular power spectrum. In both these cases the assumption of a uniform sheet of incoherent sources is adequate. However, to investigate the temporal statistics in either case, for instance to estimate averaging times, requires a fairly detailed model of the surface wave spatial field. Thus, the noise model is very much application dependent. These mechanisms and effects will be discussed in the context of experimental data.