Evaluating potential EEG-indicators for auditory attention to speech in realistic environmental noise

Abstract

The human brain is remarkably capable of perceiving relevant sounds in noisy environments but the underlying inter-play of neurophysiology and acoustics is still being investigated. Cortical processing of these sounds in the brain de-pends on attentional demand. One of the most important issues is how to identify whether a person is paying attentionto the relevant sounds or not. The aim of this study was to explore the potential of single-trial electroencephalography(EEG) indicators to distinguish the cortical representation of three sequential tasks — attentive listening to lecturesin background noise, attentive and inattentive listening to background noise alone. Three types of environmentalnoise, including multi-talker babble, fluctuating traffic and highway sounds were employed as the background duringthe first task and the stimulus during the second and third tasks. 23 healthy volunteers were exposed to these threetasks while 64-channels EEG signals were recorded. Alpha-band spectral characteristics (peak frequency and power)were investigated as potential indicators of attention and cortical inhibition. Furthermore, based on the hypothesisof self-similarity as excitation-inhibition balance, long-range temporal correlation of alpha-band activity was quanti-fied based on detrended fluctuation analysis. Finally, the hypothesis of speech envelope entrainment of brain activitymotivated to estimate the delta absolute power for investigating the attended sound. Considering the participant asa random factor, a linear mixed-effect regression was employed to model the estimated indicators as a function oflistening task, EEG channel cluster, and background noise. Strong significant differences were found that support ourhypotheses that auditory attention to speech can be observed via EEG-indicators.