Attentive tracking in human audition
Kevin J. P. Woods, Josh H. McDermott · The Journal of the Acoustical Society of America · 2016
Auditory scenes often contain multiple sound sources, but typically one is of particular interest and must be selected for further processing. This “cocktail party problem” is especially difficult when sources are similar and change over time (e.g., speakers of the same gender). To study this situation, we introduced a paradigm in which listeners attempt to follow one of two synthetic voices that vary randomly in several feature dimensions (e.g., f0, f1, and f2). Psychophysical results from this task suggest that human listeners employ a movable focus of attention that follows sources of interest through their feature space. Here we utilize this paradigm to probe the performance limits of attentive tracking. We also report that listeners can learn the abstract “shape” of an attentively tracked trajectory if it occurs repeatedly in a stimulus set. Knowledge of trajectory regularities appears to benefit the tracking of target sources, and demonstrates another way in which attentive tracking might aid our ability to hear under challenging listening conditions.