Preprocessing for noise reduction of chaotic signals

Existing algorithms for filtering noise from contaminated chaotic signals require relatively high signal-to-noise ratios (SNRs). A preprocessing method is proposed for initial noise reduction in severe SNRs. This method uses all the difference vectors between the points in a small neighborhood of the point of interest to estimate more accurately the tangent surface described by Cawley and Hsu (1992). This preprocessing method is applicable to noisy data from any nonlinear dynamical system, because it does not require knowledge of the system dynamics