Digital Signal Processing and System Theory

Talk Martin Fischer

Empirical Mode Decomposition Applied on EEG Signals in Real Time

Date: 23.11.2012, 15:00 h - 15:45 h
Room: Aquarium

Martin Fischer
CAU, Kiel, Germany


Electroencephalogram (EEG) data are often contaminated with artifacts. Several methods have been proposed to enhance the EEG recordings based on the Empirical Mode Decomposition (EMD), an adaptive data-driven technique, which decomposes non-stationary and non-linear data into a number of Intrinsic Mode Functions (IMF). Once that the IMFs are obtained, they are used for denoising and detrending purposes. This thesis presents a real-time implementation of the EMD algorithm. Overlapping windows are used to minimize the errors which arise due to the block-wise processing. Furthermore, the output of the real-time implementation is used for signal enhancement. The contaminated IMFs are identified and attenutated. The results show that the proposed techniques can succesfully suppress eye movement aritfacts and muscle artifcats, while sharp waves arising from epilepsy are preserved. A graphical user interface has been created and was added to the Kiel Real-Time Audio Toolkit. It allows the user to observe and manipulate the EMD to enhance the signals.