High-resolution temporal weighting of interaural time differences in speech

Abstract

Previous studies have shown that for high-rate click trains and low-frequency pure tones, interaural time differences (ITDs) at the onset of stimulus contribute most strongly to the overall lateralization percept (receive the largest perceptual weight). Previous studies have also shown that when these stimuli are modulated, ITDs during the rising portion of the modulation cycle receive increased perceptual weight. Baltzell, Cho, Swaminathan, and Best [(2020). J. Acoust. Soc. Am. 147, 3883–3894] measured perceptual weights for a pair of spoken words (“two” and “eight”), and found that word-initial phonemes receive larger weight than word-final phonemes, suggesting a “word-onset dominance” for speech. Generalizability of this conclusion was limited by a coarse temporal resolution and limited stimulus set. In the present study, temporal weighting functions (TWFs) were measured for four spoken words (“two,” “eight,” “six,” and “nine”). Stimuli were partitioned into 30-ms bins, ITDs were applied independently to each bin, and lateralization judgements were obtained. TWFs were derived using a hierarchical regression model. Results suggest that “word-initial” onset dominance does not generalize across words and that TWFs depend in part on acoustic changes throughout the stimulus. Two model-based predictions were generated to account for observed TWFs, but neither could fully account for the perceptual data.

Binaural consequences of speech envelope enhancement

“…The variations in the presence and size of these effects across words likely reflects the fact that EE and CE emphasize certain features of certain words (Fig. 1 ) and also that listeners place their weight on different parts of the envelope for different words ( Baltzell and Best, 2021 ).…”

“…The stimuli were four speech tokens (“two,” “six,” “eight,” and “nine”) spoken by a female talker [and used previously by Baltzell and Best (2021) ]. All stimuli were normalized to a presentation level of 60 dB SPL subsequent to processing.…”

“…Of particular relevance here, interaural time differences (ITDs) occurring at onsets in the envelope tend to dominate the overall spatial percept of modulated sounds, and the steepness of these rising portions of the envelope is known to affect the salience of ITDs ( Klein-Hennig et al , 2011 ; Laback et al , 2011 ; Dietz et al , 2013 ; Stecker, 2018 ). This also appears to be largely true for speech ( Baltzell et al , 2020 ; Baltzell and Best, 2021 ), though exactly what spectro-temporal features constitute a perceptually robust onset for speech stimuli has not been determined. It is on these bases that we might expect binaural consequences of different envelope enhancement algorithms.…”