Parameter estimation of superimposed damped sinusoids using exponential windows

• A new preprocessing technique for parameter estimation of superim-posed exponentially damped sinusoids is proposed.
• The technique is based on cyclo-stationarity, and its extension expo-nential windowing (EW).
• An ad-hoc but efficient approach for the EW parameter selection isprovided and shown to provide close to CRB performance.
• Simulation results show that the proposed method significantly im-proves the robustness to noise over two other common preprocessing techniques.
• The method is particularly effective with non-Gaussian and colored noise.

This paper presents a preprocessing technique based on exponential windowing (EW) for parameter estimation of superimposed exponentially damped sinusoids. It is shown that the EW technique significantly improves the robustness to noise over two other commonly used preprocessing techniques: subspace decomposition and higher order statistics. An ad hoc but efficient approach for the EW parameter selection is provided and shown to provide close to CRB performance.