IF-EWT-Based Framework for Classification of Motor Imagery EEG Signals
In this paper, we present a new framework for motor imagery (MI) classification in brain-computer interface (BCI) systems using electroencephalogram (EEG) signals. The proposed framework employs the iterative filtering-based empirical wavelet transform (IF-EWT) signal decomposition method to decompose EEG signals into modes. Instantaneous amplitude and instantaneous frequency are computed for each mode using Hilbert spectral analysis, and time-frequency (TF) images for each channel are then constructed. To obtain event-related desynchronization and synchronization patterns during MI tasks, the mu (8-14 Hz) and beta (16-30 Hz) bands are extracted from each channel TF image, and then combined to preserve temporal, frequency and spatial domain information for each EEG signal. These combined TF images are used as input to a convolutional neural network (CNN) for classification. The proposed framework is tested on the BCI competition IV Dataset 2b, Experiments are conducted with two-channel and threechannel configurations as input to a CNN, which achieved average classification accuracies of 94.21% and 94.39%, respectively. These results indicate that the proposed method outperforms the other existing methods.