Analysis of speech imagery using brain connectivity estimators on consonant-vowel-consonant words

Speech imagery refers to the perceptual experience of uttering speech to oneself without any articulation. In this paper, the neural correlations between brain regions associated with articulated and imagined speech processes of consonant-vowel-consonant (CVC) words are analysed using brain connectivity estimators. EEG coherence, a synchronisation parameter establishes the correlation between several cortical areas. To analyse the causal dependence, partial directed coherence (PDC) and directed transfer function (DTF) estimators are derived from multi-channel EEG data. From inter and intra hemispheric coherences it has been observed that theta, beta and gamma frequencies were dominant and words with same vowel having one consonant common have similar coherence values. Results inferred from intra-hemispheric PDC and DTF parameters show that the frontal and temporal regions of the left hemisphere are more activated for all the given speech imagery tasks. Thus, the analysis provides a significant step in understanding the neural interactions of the brain while thinking and articulating processes.