Motor Imagery Classification in EEG signals using Hybrid CNN-ISqueeze net

Motor imagery (MI) electroencephalography (MI-EEG) data are widely used in BCI systems to determine participant intention. MI classification is challenging because of several reasons, such as poor signal-to-noise ratios (SNR) and lack of excellent data. The operation of BCI devices requires an understanding of MI-EEG signals. Recent years have seen the effective use of deep learning (DL) techniques in recognizing patterns and other domains. Conversely, there aren’t many efficient DL algorithms that have been used with BCI systems. This research work presents a novel MIC scheme using the hybrid classifier. At first, the EEG signals are preprocessed by applying the band pass filtering (BPF) method. Then, features including Stockwell transform (ST), correntropy and Common Spatial Pattern (CSP) features are extracted. For MIC, a hybrid classification method that combines CNN and Improved squeeze net (ISQN) is employed in this work.