Block subband coding and time-varying filterbanks

Abstract

Adaptive block processing is a technique suitable for compressing non-stationary signals, where input blocks are first decomposed by a filterbank into subband signals. These subband signals are in turn analyzed to extract the time-varying signal parameters. We can optimize a given subband coding algorithm by introducing time-varying filterbanks, where for each input block we alter the structure of the decomposition filterbank such that a particular coding criteria is optimized. In this paper, we use time-varying filterbanks as a basis for adaptive subband coding. In particular, we analyze the relationship between adaptive block subband coding and time-varying filterbanks. We introduced a method for deriving boundary filters and entry/exit filters for the synthesis filterbanks. We show that conventional design techniques for the time-varying filterbanks exhibit many characteristics that make them less than ideal candidates for adaptive signal decomposition.