Search Papers | Poster Sessions | All Posters
Poster C67 in Poster Session C - Friday, August 9, 2024, 11:15 am – 1:15 pm, Johnson Ice Rink
Improved Modeling of EEG Responses to Natural Speech Acoustics Using Dynamic Temporal Response Functions
Jin Dou1 (), Andrew J. Anderson2, Samuel V. Norman-Haignere1, Edmund C. Lalor1; 1University of Rochester, Rochester, NY, United States of America, 2Medical College of Wisconsin, Milwaukee, WI, United States of America
Speech is central to human life. However, how the human brain converts acoustic speech into language remains incompletely understood. One common way to study this process is by deriving models that map between speech stimuli and the resulting brain responses. The temporal response function (TRF) is one such model that assumes that responses to speech are time-invariant with magnitudes that are linearly related to the amplitude of various speech features. However, such linear time-invariant assumptions are sure to be suboptimal given what is known about the brain. Here, we relax the linear time-invariant assumptions using a recently proposed dynamically warped TRF model that can modulate both the amplitude and timing of the TRF based on the current and previous values of the stimulus feature of interest. Doing this improved the ability to model EEG responses to natural speech. This improvement was driven by the dynamic TRF’s ability to account for the fact that larger acoustic onset values tended to evoke larger and earlier responses, a finding that is consistent with previous research. This study validated the efficacy of the dynamically warped TRF model and emphasizes the importance of considering the timing of brain responses to natural stimuli.
Keywords: EEG speech temporal response function nonlinear