Classifying motor imagery EEG by Empirical Mode Decomposition based on spatial-time-frequency joint analysis approach

Abstract

A novel spatial-time-frequency approach to classify the different mental task in brain computer interface was presented. A high resolution time-frequency spectral was achieved by using Empirical Mode Decomposition and Hilbert-Huang Transform, and the subject specific spatial-time-frequency joint features were extracted from the restricted spectral of multi-channel EEG recordings. A weighting synthetic classifier was built and used to identify the classes of the imaged motions The test results in four subjects showed that the classification accuracy varied between 77.0% and 95.0%, with an average of 85.9%, which suggested that the present method can achieve a reasonable performance in identifying imaged motions compared with previous methods.