Electroencephalography based imagined alphabets classification using spatial and time-domain features

Imagined speech is a neuro-paradigm that can provide an alternative communication channel for patients in a locked-in syndrome state. We have performed an experiment in which a 32 channel industry-standard electroencephalography (EEG) device was used to record 26 imagined English alphabets from 13 subjects. We denoised the imagined signals by discrete wavelet transform and extracted the spatial filters by common spatial pattern method, and time-domain features. Spatial features when classified with linear support vector machine, and time-domain features classified by random forest gave the best results. Alpha, beta, and theta bands could classify imagined alphabets better than other bands and had average classification accuracies of 88.59%, 87.39%, and 88.97%, respectively by using spatial features and 81.88%, 76.72%, and 79.25%, respectively, by time-domain features. The grand average accuracies of all the 26 alphabets in six EEG frequency bands was found to be 77.97% in a subject independent binary classification framework.