modelling and deconvolution of nonstationary acoustic signals from moving sources using a convariance equivalent formulation

This paper is concerned with the modelling and 'stationarisation' of a class of nonstationary random processes. Such processes arise as the sound perceived by an observer due to moving acoustic sources. The sources are assumed to be a collection of monopoles which are considered to emit random signals which are stationary in the frames of reference of the sources. The random processes involved are assumed to admit a state space shaping filter formulation and the concept of covariance equivalence is used to describe the temporal behaviour of the perceived signals as white noise excited time varying linear filters. This allows the evolutionary (auto and cross) spectral densities to be computed. The stationarisation of this class of non-stationary signals is achieved by using an optimum control based deconvolution procedure.