Learning Common Time-Frequency-Spatial Patterns for Motor Imagery Classification

The common spatial patterns (CSP) algorithm is the most popular filtering method applied to extract electroencephalogram (EEG) features for motor imagery (MI) based brain-computer interface (BCI) systems. effectiveness of CSP depends on optimal selection frequency band and time window from EEG. Many algorithms have been designed optimize CSP, while few seek window. This study proposes a novel framework, termed time-frequency-spatial (CTFSP), sparse multi-band filtered EEG data in multiple windows. Specifically, whole MI period first segmented into subseries using sliding approach. Then, are extracted bands each Finally, support vector machine (SVM) classifiers with Radial Basis Function (RBF) kernel trained identify tasks voting result these determines final output BCI. applies proposed CTFSP three public datasets (BCI competition III dataset IVa, BCI IIIa, IV 1) validate its effectiveness, compared against several other state-of-the-art methods. experimental results demonstrate that promising candidate improving performance MI-BCI