Abstract
AbstractIn this study, we use electroencephalography (EEG) recordings to perform absolute auditory attention detection (aAAD), i.e., determine whether a subject is actively listening to a presented speech stimulus or not. More precisely, we aim to discriminate between an active listening condition, and a distractor condition where subjects passively listen to the speech stimulus while performing another cognitive task. To this end, we re-use an existing EEG dataset where the subjects watch a silent movie as a distractor condition, and introduce a new EEG dataset with two other distractor conditions (silently reading a text and performing arithmetic exercises). We focus on two EEG features, namely neural envelope tracking (NET) and spectral entropy (SE). We find significantly higher NET and lower SE in the active listening condition compared to the distractor conditions, which for the SE is the reverse of what was previously found for an active listening versus passive listening condition (without any distractors). In addition, aAAD is used in the context of a selective auditory attention decoding (sAAD) task, where the goal is to decode to which of two competing speakers the subject is attending, which is a core task in the context of so-called neuro-steered hearing devices. We show that evaluating sAAD performance only on segments of active listening improves sAAD performance when detecting these active listening segments as having higher NET, whereas the reverse trend is observed when detecting these segments as having lower SE. We conclude that NET is a more reliable metric for aAAD as it is consistently higher for the active listening condition, whereas the relation of the SE between the active listening and passive listening conditions seems to depend on the nature of the distractor task. Consequently, NET shows the most promise for aAAD and to detect auditory inattentive segments in neuro-steered hearing devices.
Publisher
Cold Spring Harbor Laboratory