Applying a Novel ICA Algorithm to Single-trial Flat EEG Analysis

摘要

As a typical example of lost consciousness, studying on the flat EEG and brain death judgment criteria are the most important social issues. Due to the lower-power activities, artifacts such as heart action and environmental noise are contaminated in the EEG signals, therefore, the judgment of brain death based on the flat EEG wave is always debatable in clinic. In this report, we present a robust approach for decomposing unaveraged single-trial EEG data. Our approach has two procedures. In the first step, a PCA-like pre-whitening technique with an additive noise reduction technique is presented. In the second step, a robust nonlinear function derived by the parameterized t-distribution model is applied to decompose the mixtures of sub-Gaussian and super-Gaussian source components.