We describe a novel algorithm to estimate the pulse pressure variation index (PPV) from arterial blood pressure signals (ABP). PPV has been shown to be one of the best predictors of fluid responsiveness in mechanically ventilated subjects. Our PPV algorithm uses a non-linear technique for envelope estimation, eliminating the need for automatic beat detection. Additionally, the algorithm makes use of nonparametric spectral techniques to extract the respiratory rate, and a median filter for artifact removal. The algorithm was validated against the continuous PPV output obtained from the commercially available PiCCOreg system and gold standard expert PPV manual annotations. The data consists of ABP taken from subjects who experienced rapid changes in hemodynamics. This data comprised over six hours of continuous ABP monitoring.