Brain monitoring of distraction from speech in noisy context using EEG

Speech is one of the most important forms of communication, yet it is often embedded in background noise, such as babble, traffic, and other environmental sounds. This background impairs speech intelligibility but can also distract attention away from the narrative. The main goal of this study is to identify neurological EEG biomarkers that may indicate whether participants are paying attention to the information that is presented, or not. Prior work has used trained computer models to identify which of two interfering speech segments is attended to. The current research focusses on non-speech background sound and in particular on the possibility to identify distraction by salient events such as a phone ringing, a car horn’s honk, or an emergency siren.

Five-minute meaningful speech fragments were presented in a very low level of pink noise and in three types of background noise. The level of the background sound was low enough not to cause energetic masking. Participants were instructed to pay attention to the lecture and were informed they would be questioned about its content. The background sounds were also presented separately during a resting period where participants were asked not to pay attention to the sound. In total 23 participants were exposed to the stimuli while their 64-channels EEG was recorded. For exclusion purposes, a full battery of audiological tests was also performed. After cleaning artifacts, the single-trial EEG was analyzed using a wavelet spectrogram. The overall spectral power as well as the response evoked by assumed distractor (e.g. phone ringing) were assessed.

In agreement with previous research, significant differences were found between participants. However, there were also some consistent trends. When comparing the attentive listening to speech in background noise to inattentive exposure to background: (a) an increase in low-frequency fluctuating power (especially in the frontal channels) was observed which might be attributed to an amplitude following response; (b) an increase in gamma power in F4 (right) in comparison with F3 (left) occurs, which could be related to linguistic processing of speech. Based on the hypothesis of increased inhibition and gating out of disturbing events during attentive as well as inattentive listening, an increase in frontal alpha power could be expected during phone ringing. This was only confirmed in the background-only samples. Between people, more alpha power in the frontal area seems related to lower amplitude following responses to the disturbing event.