EEG signal classification with super-Dirichlet mixture model

Classification of the Electroencephalogram (EEG) signal is a challengeable task in the brain-computer interface systems. The marginalized discrete wavelet transform (mDWT) coefficients extracted from the EEG signals have been frequently used in researches since they reveal features related to the transient nature of the signals. To improve the classification performance based on the mDWT coefficients, we propose a new classification method by utilizing the nonnegative and sum-to-one properties of the mDWT coefficients. To this end, the distribution of the mDWT coefficients is modeled by the Dirichlet distribution and the distribution of the mDWT coefficients from more than one channels is described by a super-Dirichlet mixture model (SDMM). The Fisher ratio and the generalization error estimation are applied to select relevant channels, respectively. Compared to the state-of-the-art support vector machine (SVM) based classifier, the SDMM based classifier performs more stable and shows a promising improvement, with both channel selection strategies.