Time Domain Analysis of EEG to Classify Imagined Speech

Electroencephalography (EEG) finds variety of uses in the fields ranging from medicine to research. EEG has long been used to study the different responses of the brain. In this paper, EEG has been applied to study the imagined vowel sounds. An algorithm is developed to differentiate three classes of imagined vowel sounds namely /a/, /u/, and ‘rest or no action’ in pairwise manner. The algorithm is tested on three subjects S1, S2, and S3 and high performance is achieved. With classification accuracy ranging from 85 to 100 %, the algorithm shows the potential to be used in Brain Computer Interfaces (BCIs) and synthetic telepathy systems. High classification performance is obtained. Sensitivity ranges from 90 to 100 %. Specificity ranges from 80 to 100 %. Positive predictive value ranges from 81.82 to 100 %. Negative predictive value ranges from 88.89 to 100 %.