Binaural prediction of speech intelligibility in reverberant rooms with multiple noise sources

Abstract

When speech is in competition with interfering sources in rooms, monaural indicators of intelligibility fail to take account of the listener's abilities to separate target speech from interfering sounds using the binaural system. In order to incorporate these segregation abilities and their susceptibility to reverberation, Lavandier and Culling [J. Acoust. Soc. Am. 127, 387-399 (2010)] proposed a model which combines effects of better-ear listening and binaural unmasking. A computationally efficient version of this model is evaluated here under more realistic conditions that include head shadow, multiple stationary noise sources, and real-room acoustics. Three experiments are presented in which speech reception thresholds were measured in the presence of one to three interferers using real-room listening over headphones, simulated by convolving anechoic stimuli with binaural room impulse-responses measured with dummy-head transducers in five rooms. Without fitting any parameter of the model, there was close correspondence between measured and predicted differences in threshold across all tested conditions. The model's components of better-ear listening and binaural unmasking were validated both in isolation and in combination. The computational efficiency of this prediction method allows the generation of complex "intelligibility maps" from room designs.

Binaural speech intelligibility in rooms with variations in spatial location of sources and modulation depth of noise interferers

“…The location of the listener was fixed in each room and the talker was moved around to different bearings and distances. Full information about the measurements can be found in Lavandier et al (2012). Only a small subset of these BRIRs were used in the present study (see the specific design of each experiment for more details).…”

“…SRTs were measured for stationary noises and noises modulated by the envelopes of speech signals comprising one, two, or four voices. The interferers were tested at three distances (65, 125, and 500 cm) in front of the listener in a lecture hall (Lavandier et al, 2012). The distances corresponded to three levels of reverberation covering a wide range of C50 (early-to-late energy ratio; ISO 3382, 1997), varying from 21 dB at 65 cm to 9 dB at 5 m. The intermediate level of 16 dB at 1.25 m was chosen to see if the ability to exploit interferer modulations would be affected by even moderate reverberation.…”

“…Binaural models have been proposed to predict spatial unmasking from a stationary-noise source in rooms (Beutelmann and Brand, 2006;Lavandier and Culling, 2010). Lavandier and colleagues recently revised their model to be able to consider multiple interferers (Jelfs et al, 2011;Lavandier et al, 2012); while Beutelmann and colleagues adapted their model so that it could be applied to a non-stationary noise (Beutelmann et al, 2010). This last implementation relied on the assumption that spatial unmasking and dip listening were independent perceptual mechanisms.…”

Head orientation benefit to speech intelligibility in noise for cochlear implant users and in realistic listening conditions

“…The model uses binaural-room impulse responses in order to reflect the impact of reverberation, when present. The Jelfs et al model has been validated against a wide variety of SRT data Jelfs et al, 2011;Lavandier et al, 2012), predicting the level of SRM in different spatial configurations with different numbers of masking noises and in different levels of reverberation. Increased SRM was predicted when listeners faced a location between the speech source and a single interfering noise source.…”

Physiology, Psychoacoustics and Cognition in Normal and Impaired Hearing

“…This would make previously undetectable speech elements audible and enable them to contribute to speech intelligibility. More recent models of speech intelligibility also make use of the concept of within-band improvement of SNR to model the spatial release from masking (e.g., Beutelmann et al 2010;Lavandier et al 2012;Wan et al 2014).…”