Models for non-stationary reverberation noise

Abstract

The random phasor approach to reverberation noise gives Gaussian statistics for intensity and is stationary in time and space. We show that fundamentally different things happen when the noise is defined by a superposition of random propagating pulses that have width. We calculate the nonstationary autocorrelation function of such a model and describe how it evolves in time and space in a dispersive medium. Furthermore, we give criteria for when a nonstationary autocorrelation function is locally stationary in time and/or space. We also consider how and under what conditions the noise evolves to the stationary case given by the ransom phasor sum. [Work supported by ONR.]