Research on emg artifact suppression method for EEG signals based on 1D-CNN-LSTM hybrid model

Abstract

Electroencephalography (EEG), as the core physiological signal reflecting the electrical activity of brain neurons, is susceptible to muscle activity interference during acquisition, generating Electromyography (EMG) artefacts that prove difficult to separate using conventional methods. Therefore, this paper proposes a hybrid 1D-CNN + LSTM framework incorporated with a mixed loss function. This approach enables precise suppression of EMG artefacts while retaining valid EEG components, realizing efficient and distortion-free purification of EEG signals.