MULTI-CNN FEATURE FUSION FOR EFFICIENT EEG CLASSIFICATION

Recently, motor imagery (MI) signals have been used in BCI systems for disabled persons, controlling robots or wheelchairs or even driving cars. Hence, researchers are using machine learning and deep learning techniques for decoding MI signals. The properties of EEG signals like a low signal to noise ratio and its dynamic nature make it more complex and harder to decode. Research has shown that EEG has both spatial and temporal characteristics which can be exploited by deep learning models like convolution neural networks (CNN). This paper shows that the multilevel CNN model can extract dynamic correlations from EEG MI data. Multilevel CNN models are fused using autoencoder to extract the best features for EEG data which help improve MI decoding accuracy. The proposed fusion methods give a good performance for EEG decoding when tested for within-subject trials as well as cross trials. The results are compared with state-of-the art-techniques on public EEG dataset BCI competition IV 2a. Novel cross encoding technique is also proposed which helps us achieve big improvement in cross-subject EEG decoding.