An Auditory Oddball Based Brain-Computer Interface System Using Multivariate EMD

摘要

A brain-computer interface (BCI) is a communication system that allows users to act on their environment by using only brain-activity. This paper presents a novel design of the auditory oddball task based brain-computer interface (BCI) system. The subject is presented with a stimulus presentation paradigm in which low-probability auditory targets are mixed with high-probability ones. In the data analysis, we employ a novel algorithm based on multivariate empirical mode decomposition that is used to extract informative brain activity features through thirteen electrodes' recorded signal of each single electroencephalogram (EEG) trial. Comparing to the result of arithmetic mean of all trials, auditory topography of peak latencies of the evoked event-related potential (ERP) demonstrated that the proposed algorithm is efficient for the detection of P300 or P100 component of the ERP in the subject's EEG. As a result we have found new ways to process EEG signals to improve detection for a P100 and P300 based BCI system.