IMPROVED NONLINEAR NOISE ESTIMATION IN EEG SIGNAL BASED ON VOLTERRA FILTERS

Abstract

Recognition and compensation of undesired nonlinearity is one of the important subjects in the field of digital signal processing. The Volterra model is widely used for nonlinearity identification in practical applications. The current tendency in the digital systems design is the identification and compensation of unwanted nonlinearities. In this paper, we employed a nonlinear noise estimation approach for electroencephalogram (EEG) signal based on a combination of linear predictive coding (LPC) and Volterra filter that is a new and good way to estimate noise in EEG signal. We initially used LPC filter to estimate the noise present in EEG signal (correlated and uncorrelated noise) plus the uncorrelated portion of the signal (the part of the signal that has no linear relation to its past samples). After that, we employed nonlinear Volterra model to estimate the existing noise in EEG signal (correlated and uncorrelated noise). We show that by employing the cascade of LPC and Volterra filter, we can considerably improve the signal-to-noise ratio (SNR) in EEG signal by the ratio of at least 1.94. Also, we compared the simulation results to the case where we used just Volterra filter. In comparison with just Volterra filter, we have a significant increase in the SNR.