Two-directional discriminative spatial patterns for movement-related EEG classification

Abstract

The discriminative spatial patterns (DSP) approach is a classical and effective feature extraction technique for single-trial EEG classification in movement task. However, it utilizes only the spatial information of EEG signals. On the other hand, the temporal information plays an important role in EEG sequences with high temporal resolution. In this paper, we propose a novel method which uses DSP to simultaneously filter the temporal and spatial dimensions to extract spatiotemporal information. Experimental results of single-trial EEG classification demonstrate that the proposed 2DDSP method obtains better classification accuracy than DSP, while the dimension of the feature matrix produced by 2DDSP is less than that of DSP.