A self-tuning adaptive trend extraction method for process monitoring and diagnosis

► We develop a method to extract on-line qualitative trends on signals. ► The method is self tuned and its tuning parameters are adapted on-line in function of the estimated level of background noise. ► An abrupt change detector is added to the algorithm to detect with accuracy the step changes occurring on a signal. ► The performances are analysed on simulated data and on real physiological data and are shown to be better than the former non-adaptive method.