In this paper, an on-line segmentation algorithm, developed to preprocess continuously monitored data in Intensive Care Units, in a purpose of alarm filtering, is presented.

The algorithm splits the signal monitored into linear affine functions of various lengths and determines on line when a new segment must be calculated. The delay of detection of a new linear function depends on the importance of the change : the more important the change, the quicker the detection. The algorithm provides a good filtering of the data, without distortion, on simulated data as well as on clinical data recorded on patients admitted in ICU. The information returned by the algorithm can be used to extract on line information on the signal, like the trend of the signal, at short or long term. This useful information could be used in “more intelligent” alarm systems, like knowledge based systems.