Consistent estimation of signal parameters in non-stationary noise

This paper addresses the problem of estimating the parameters of a deterministic signal in non-stationary, white, Gaussian noise. It is proposed to model the time-varying white Gaussian noise with an unknown deterministic variance sequence that changes every sample. While making relatively few assumptions on the non-stationarity of the noise, this type of modeling gives rise to different difficulties. We identify the resulting difficulties and discuss possible solutions.