Real-time Empirical Mode Decomposition for EEG signal enhancement

Abstract

Electroencephalography (EEG) recordings are used for brain research. However, in most cases, the recordings not only contain brain waves, but also artifacts of physiological or technical origins. A recent approach used for signal enhancement is Empirical Mode Decomposition (EMD), an adaptive data-driven technique which decomposes non-stationary data into so-called Intrinsic Mode Functions (IMFs). Once the IMFs are obtained, they can be used for denoising and detrending purposes. This paper presents a real-time implementation of an EMD-based signal enhancement scheme. The proposed implementation is used for removing noise, for suppressing muscle artifacts, and for detrending EEG signals in an automatic manner and in real-time. The proposed algorithm is demonstrated by application to a simulated and a real EEG data set from an epilepsy patient. Moreover, by visual inspection and in a quantitative manner, it is shown that after the EMD in real-time, the EEG signals are enhanced.