IEEE Project Abstract

Among various physiological signal acquisition methods for the study of human brain,EEG (Electroencephalography) is more effective. EEG provides a convenient, non-intrusive and accurateway of capturing brain signals in multiple channels at fine temporal resolution. We propose an ensemblelearning algorithm for automatically computing the most discriminative subset of EEG channels forinternal emotion recognition. Our method describes an EEG channel using kernel-based representationscomputed from the training EEG recordings. For ensemble learning, we formulate a graph embedding lineardiscriminant objective function using the kernel representations. The objective function is efficiently solvedvia sparse non-negative Principal Component Analysis (PCA) and the final classifier is learned using thesparse projection coefficients. Our algorithm is useful in reducing the amount of data while improvingcomputational efficiency and classification accuracy at the same time. Experiments on publicly availableEEG dataset demonstrate the superiority of the proposed algorithm over the compared methods.