Hilbert-Huang Time-Frequency Analysis of Motor Imagery EEG Data for Brain-Computer Interfaces

摘要

Brain-computer interface (BCI) is a technology that provides a non-muscular communication channel between a brain and the outside world. Imagination of left and right hand movements results in spatially distinct brain activation patterns that can be used as control signals for the BCI. Motor imagery (MI) results in the attenuation (event related desyn-chronization, ERD) or enhancement (event related synchronization, ERS) of amplitude in a certain frequency band of electroencephalogram (EEG). This frequency band can vary between different participants. Therefore time-frequency (TF) analysis is performed in order to extract interesting features from EEG. A simple way of performing TF analysis is by using band power features. In this paper we investigate the perspective of Hilbert-Huang transform (HHT) for extracting TF information used for MI classification. HHT is a method that allows calculation of instantaneous frequency and amplitude of the signal. It does that by decomposing the signal into components for which these parameters can be calculated by means of Hilbert transform. We compare classification accuracy of simple band power features and features obtained by means of HHT on BCI competition TV dataset 2b.