For speech interferers, BMLD unmasking acts within spectro-temporal units rather than frequency bands

Evidence

Whereas Levitt and Rabiner ( 1967 ) proposed their model for speech in stationary noise, we are interested in speech interferers. Since speech shows strong modulations both in time and frequency, the level of the interferer will vary. We will therefore consider BMLD effects at the level of spectro-temporal regions rather than frequency channels. Thus we extend the original hypothesis and postulate that a BMLD effect might take place at the level of individual spectro-temporal units and that this detection advantage would be responsible for the spatially improved speech intelligibility. — Dijk, Pim van, Physiology, Psychoacoustics and…, 1623