We can learn about the neurobiology of speech processing through recording the brain activity of a person listening to speech through electroencephalography (EEG). The EEG responses then contain features of the speech signal, for instance its envelope as well as a frequency-following response to the temporal fine structure (speech-FFR), albeit at a low signal-to-noise ratio.
The recent ICASSP 2023 ‘Auditory EEG Decoding’ Signal Processing Grand Challenge addresses the issue of correctly relating an EEG signal to a speech stimulus despite the noise in the EEG signal. Our contribution won this challenge through an approach that combines the response to the speech envelope with the speech-FFR.
M. Thornton, D. Mandic, T. Reichenbach,
Relating EEG recordings to speech using envelope tracking and the speech-FFR,
ICASSP 2023 – 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2023), 1-2. [bioarxiv] [pdf]