Adaptive estimation of EEG for subject-specific reactive band identification and improved ERD detection.

The event-related desynchronization (ERD) is a magnitude decrease phenomenon which can be found in electroencephalogram (EEG) mu-rhythm in a certain narrow frequency band (reactive band) during different sensorimotor tasks and stimuli. The success of ERD detection depends on proper identification of subject specific reactive band. An adaptive algorithm band limited multiple Fourier linear combiner (BMFLC) is employed in this paper for identification of subject specific reactive band for real-time ERD detection. With the time-frequency mapping obtained with BMFLC, a procedure is formulated for reactive band identification. Improved classification is obtained by applying this method to a standard BCI data set compared to traditional ERD detection methods. Study conducted with 8 subjects drawn from BCI Competition IV data set show a 22% increase in ERD and 10% improvement in classification with the proposed method compared to standard ERD based classification.