A Low-Computational Complexity System for EEG Signals Compression and Classification

Brain-computer interface (BCI) has emerged as a promising technology to explore physiological activities and functional states of the human brain. This paper presents an Electroencephalogram (EEG) signals processing system, which consists of the compression process, reconstruction process and the classification process. The EEG signals is reconstructed with 5.318±0.08 root mean square error and 0.881±0.06 structural similarity index in short CPU time. The experiments on the two classes motor imagery EEG signals reaches up to 92% accuracy with the proposed feature extraction method. Compared with prior works, the proposed work is able to achieve better performance with high fidelity and low computational complexity.