Abstract
Speech Imagery is one form of Mental Imagery which refers to imagining of speaking a word or sentence to oneself silently in the mind without any articulation movement. Speech Imagery concept will be useful for patients with neurological disorders for decoding their thoughts and possibly convert their thought to speech through some external medium. In this paper Electroencephalography (EEG) signals were acquired from two healthy subjects while speaking and imagining of speaking five consonant-vowel-consonant (CVC) words in multiple trials of different time frames. Each trial was carried out for five times during both the tasks. The acquired EEG signals wereanalyzed in frequency domain using Fast Fourier Transform and the power spectral densities for each band were derived. Based on the derived absolute power, relative spectral powers were computed for each EEG frequency band during speaking and imagining of speaking. Relative powers for both inter and intra hemispheric frontal and temporal electrodes were analyzed as these electrodes are located near Broca's and Wernicke's area. From theresults it was inferred that relative power of theta band was found to be high while imagining of speaking a CVC word and similarly relative power of alpha band was found to be dominant while speaking a CVC word. It has also been observed that averaged relative power of left hemispheric frontal and temporal electrodes was increasing progressively as the resting time for the task is increased. Comparatively the averaged relative power was found to be higher than the relative power of a single trial. This work will be clinically useful for speech impaired people as it is feasible to recognize the word from these signal processing methods.