Individual variability and similarity in horizontal directivity patterns for speech

Abstract

Speech directivity patterns are highly complex and temporally dynamic during speech production, with notable variability across individuals. In this paper, we report some of the complexities observed in anechoic horizontal speech directivity data obtained from standing talkers, including variability and consistencies observed in speech directivity patterns across talkers. Although speech generally becomes more directional towards 0º azimuth as frequency increases, it is clear that the relationship between frequency and directionality is nonmonotonic. Our data indicate this relationship shows some consistency across talkers. For example, peak directionality towards 0º tends to occur around 7-9 kHz, with a decrease in directionality at higher frequencies. On the other hand, due to acoustic lobing, minimum directionality towards 0º and corresponding maximum directionality towards 90º tends to occur around 700-800 Hz. Whether these particular phenomena have perceptual consequences for listeners in complex auditory scenes is an intriguing but unresolved question. However, previously published perceptual data do indicate talker effects on listener detection of talker head orientation, suggesting individual variability in speech directivity patterns may be perceptually relevant.