Sinc-Based Convolutional Neural Networks for EEG-BCI-Based Motor Imagery Classification

Brain-Computer Interfaces (BCI) based on motor imagery translate mental images recognized from the electroencephalogram (EEG) to control commands. EEG patterns of different imagination tasks, e.g. hand and foot movements, are effectively classified with machine learning techniques using band power features. Recently, also Convolutional Neural Networks (CNNs) that learn both effective features classifiers simultaneously raw data have been applied. However, CNNs two major drawbacks: (i) they a very large number parameters, which thus requires training examples; (ii) not designed explicitly in frequency domain. To overcome these limitations, this work we introduce Sinc-EEGNet, lightweight CNN architecture combines learnable band-pass depthwise convolutional filters. Experimental results obtained publicly available BCI Competition IV Dataset 2a show our approach outperforms reference methods terms classification accuracy.