Identification of Multi-Channel Simulated Auditory Event-Related Potentials using a Combination of Principal Component Analysis and Kalman Filtering

Abstract

The aim of this paper to develop a new method for separating auditory event-related potentials (ERP) signal from artefacts or noise. In experimental conditions, ERPs can be approximated by weighted sums of Principal Component Analysis (PCA) basis signals calculated from clean data. Projection of measured signals onto the PCA subspace significantly decreases noise. Furthermore, Kalman filtering has been used to optimize the combining of the PCA filtered signal with an a priori expected ERP. The main strength of the proposed algorithm arises from manipulating a priori cross-channel information in the form of a PCA weight covariance matrix. Here, the implementation of the method has been quantified using synthetic multi-channel ERP signals to which known amounts of synthetic noise is added to all the channels. The use of synthetic data means and signal and noise are known and so signal-to-noise enhancement may be quantified. For a wide range of initial SNRs, PCA filtering increases SNR by 10 dB and Kalman filtering yields an additional 10 dB improvement.