Multi frequency band fusion method for EEG signal classification

Abstract This paper proposes a novel convolutional neural network (CNN) fusion method for electroencephalography (EEG) motor imagery (MI) signal classification. The is named MFBF, which stands multifrequency band fusion. MFBF relies on filtering the input with different frequency bands and feeding each to duplicate of CNN model; then, all duplicates are concatenated form model. also introduces second release Coleeg software, used evaluation. has advantage flexibility choosing any model number bands. In experimental evaluation, CNN1D three were CNN1D_MFBF model, it was evaluated against EEGNet_fusion datasets, are: Physionet, BCI competition IV-2a, dataset from Hungarian Academy Sciences Research Centre Natural (MTA-TTK). had comparable or better accuracy results less than one-fifth training time, significant proposed method.