A coherence analysis of EEG response to speech modulated by m-sequence
Hiroshige Takeichi, Sachiko Koyama, Brett L. Foster, David T. J. Liley · Swinburne figshare (Swinburne University of Technology) · 2008
We have developed a technique for assessment of natural verbal comprehension, using electroencephalography (EEG) and m-sequence modulation (Takeichi et al., Neuroscience Research, 57, 314-318). The technique is superior to the conventional averaging methods in that it is less susceptible to measurement artifacts, efficient without repeated presentations for signal averaging, and suitable for non-time-locked continuous mental processes. Scalp EEGs were recorded from eight right-handed Japanese speakers while they were listening to 51s-long news, in Japanese and Spanish, with and without modulation in amplitude timed by an m-sequence (a binary pseudorandom sequence). Correlation functions were computed between the m-sequence and the EEG. Then, independent component analysis (ICA, AMUSE, delta=1) was applied to the correlation functions to find critical components to discriminate between comprehensible and incomprehensible stimuli. In the present study, we focused on the EEG coherence between recording electrodes. The original EEGs were convolved with the Fourier transform of the critical ICA component function, in order to maintain only the critical component. Coherence values were computed for each electrode pair and individual frequency bands including theta, alpha, beta, gamma, and were averaged across subjects for each frequency band. The coherence value for the Japanese speech without modulation increased between T7 (left temporal) and the other electrodes in the alpha (4-13Hz) band. For the modulated Japanese coherence increased between T7 (left temporal) and the other electrodes, and F8 (right frontal) with other electrodes in the high-gamma (70-100Hz) band. Such coherence changes were not found for either nonmodulated or modulated Spanish speech.