A Refined Method of Robust Smoothing

Nonlinear data smoothers (filters) have the drawback that too many local peaks and troughs in a data sequence may be preserved. If Gaussian assumptions are not met, linear smoothers do not offer a desirable alternative. A refined method of smoothing out local peaks and troughs, while retaining the broad ones, is proposed. When used for signal recovery from data sequences contaminated with noise, this procedure, termed splicing, appears superior to other methods considered. Application to a real data sequence is presented as an illustration of the technique.