Median filtering algorithms for multichannel detectors

Abstract

Particle detectors of worldwide networks are continuously measuring various secondary particle fluxes incident on Earth surface. At the Aragats Space Environmental Center (ASEC), the data of 12 cosmic ray particle detectors with a total of ∼280 measuring channels (count rates of electrons, muons and neutrons channels) are sent each minute via wireless bridges to a MySQL database. These time series are used for the different tasks of off-line physical analysis and for online forewarning services. Usually long time series contain several types of errors (gaps due to failures of high or low voltage power supply, spurious spikes due to radio interferences, abrupt changes of mean values of several channels or/and slowly trends in mean values due to aging of electronics components, etc.). To avoid erroneous physical inference and false alarms of alerting systems we introduce offline and online filters to “purify” multiple time-series. In the presented paper we classify possible mistakes in time series and introduce median filtering algorithms for online and off-line “purification” of multiple time-series.