EEG data compression with source coding techniques

Abstract

A data compression algorithm for the EEG, derived from the adaptive pulse code modulation scheme, is described where the consecutively computed differences are coded by passing them through a quantizer possessing only a few levels; the range of these levels is adapted to local signal statistics. Three different versions of the algorithm with data reduction up to 75% are presented. The system was validated using several multichannel-routine EEG recordings with both visual evaluation and computation of signal-to-noise ratios.