12th Speech in Noise Workshop, 9-10 January 2020, Toulouse, FR

Listening effort and oesophageal speech: An EEG study

Sneha Raman(a)
AHOLAB Signal Processing Laboratory, University of the Basque Country (UPV/EHU), Bilbao, Spain

Axel Winneke
Fraunhofer Institute for Digital Media Technology, Branch Hearing, Speech and Audio Technology, Oldenburg, Germany

Inma Hernaez, Eva Navas
AHOLAB Signal Processing Laboratory, University of the Basque Country (UPV/EHU), Bilbao, Spain

(a) Presenting

Listening effort (LE) is being increasingly investigated in several ways as it helps us understand not just how much of the information was understood correctly, but also how taxing it was to listen to the message. Electroencephalography (EEG) has been used to investigate brain activity related to LE. However, this has been done mostly for speech in noise (SPiN) and/or in the context of hearing impairment. EEG experiments for pathological speech — which is intrinsically noisy — have not been explored extensively.

We performed an EEG experiment with participants with normal hearing (N=12) to look at the differences in brain activity when listening to healthy and impaired speech. The impaired speech we have used is oesophageal speech (OS), which is one of the speech production mechanisms adopted following a laryngectomy (removal of larynx). OS — generated using the vibrations of the oesophagus — lacks fundamental frequency and contains unwanted artefacts such as swallowing sounds, which affects the intelligibility, rhythm and prosody of speech. These alterations make OS more effortful to listen to compared to healthy speech. Participants listened to sentences in healthy and impaired speech and were asked to subjectively rate the perceived LE on a 14-point scale ranging from ‘no effort’ to ‘extreme effort’. A multichannel continuous EEG was recorded while the task was performed. We hypothesised differences in LE when listening to healthy and impaired speech and that these differences should also be reflected in the brain activity. As expected, the subjective data revealed significantly more LE associated with impaired as compared to healthy speech.

The focus for the EEG data analysis was placed on activity in the alpha and delta frequency band. Alpha (8-12 Hz) power has been found to be related to LE for SPiN and suggested to reflect the suppression of task-irrelevant information. Increased delta activity (0.5-3 Hz) in the frontal region has been found to increase with increasing levels of concentration in tasks such as mental calculations. In our EEG data, no significant differences between the two conditions were found with respect to alpha power in the parietal region. However, there was a significant increase in the delta band power when listening to impaired as compared to healthy speech, particularly in frontal regions. These results suggest that the difficulty in listening to impaired speech may not be attributed to requiring suppression of distractors but with needing increased concentration to perceive it.

Last modified 2020-01-06 19:23:55