A Sparse Multiclass Motor Imagery EEG Classification Using 1D-ConvResNet

Multiclass motor imagery classification is essential for brain–computer interface systems such as prosthetic arms. The compressive sensing of EEG helps classify brain signals in real-time, which necessary a BCI system. However, limited, despite its flexibility and data efficiency, because sparsity high computational cost reconstructing signals. Although the constraint has been addressed through neural networks, signal reconstruction remains slow, increases to further. Therefore, we propose 1D-Convolutional Residual Network that classifies features compressed (sparse) domain without signal. First, extract only wavelet (energy entropy) from raw epochs construct dictionary. Next, given test based on sparse representation proposed method computationally inexpensive, fast, accuracy it uses single feature preprocessing. trained, validated, tested using multiclass 109 subjects PhysioNet database. results demonstrate outperforms state-of-the-art classifiers with 96.6% accuracy.