Abstract
In this paper a new method for localization of sources of brain rhythmic activity is presented. The empirical mode decomposition (EMD) method is applied to an appropriate channel and one of the extracted intrinsic mode functions (IMFs) is selected as a reference signal for one of the brain rhythms. Then the spatial notch filter which is a constrained spatial filter based on a reference is applied to find the location of the desired rhythmic source. The use of EMD which is fully adaptive and data-driven method for analyzing non-stationary and nonlinear time series along with the recently developed spatial filter is a powerful method for localization of different rhythms in different frequency bands inside the brain. The method is applied on the simulated data and real BCI database. The results validate the effectiveness of the proposed method for localization of sources with different time-frequency signatures. © EURASIP, 2009.