Rank Order Filters for Time-Stamped Signals

The emerging availability of precision time synchronization, such as that provided by IEEE 1588 Precision Time Protocol, make it desirable to have signal processing methods available for signals that are sampled at non-uniform intervals but have the sample times accurately recorded, or time-stamped. Rank order filters, including median filters, are well known in discrete time signal processing. Such filters are used for such purposes as signal de-noising without ringing and robust esimation of signal mimimum and maximum values. The question considered in this paper, then, is: Are there efficient analogues in the domain of time stamped signals for rank order filters ? The answer is yes. This paper shows how to transfer a theory from the literature concerning rank order filters for analog signals to the domain of time-stamped signals, discusses how to implement the resulting rank order filters efficiently, and shows the results on actual time-stamped, measured data.