Partition-based filters

Abstract

In this paper we have introduced and analyzed a new class of adaptive nonlinear filters referred to as partition-based linear (Pl) filters. The operation of these filters is based on partitioning the R/sup N/ observation space defined by a size N moving observation window. Specifically, scaler quantization and vector quantization (VQ) have been used as useful examples to illustrate the concept of partitioning the observation space. Each partition is assigned a corresponding set of filter weights. Given that an observation vector lies in a certain partition, the filter uses the corresponding set of weights and forms an estimate by taking a linear combinations of the observation samples. Hence, the name partition-linear filters. Simulations include a novel approach to estimating response-to-response variations in evoked potentials (EP), buried in the on-going electroencephalogram (EEG).