Applying Outlier Detection and Independent Component Analysis for Compressed Sensing EEG Measurement Framework

We previously proposed a framework that applies independent component analysis (ICA) after random undersampling to compress EEG signals with artifacts. In the previous framework, compressed signals are decomposed into several independent EEG signal components, and all components that include artifacts are set to zero in order to reconstruct the compressed EEG measurement signal without artifacts. In this paper, we propose a framework using outlier detection-ICA instead of ICA for efficiently removing only the artifacts. Herein, the performance of the proposed framework is evaluated using raw EEG signals with an ocular artifact pseudo-model. Compared with previous framework, the addition of outlier detection lowers the normalized mean square error and preserves the EEG signals with a high accuracy.