Inteligibilidad
Contents lists available at ScienceDirect
Hearing Research
journal homepage: www.elsevier.com/locate/heares
Research paper
Revision and validation of a binaural model for speech intelligibility in noise
Sam Jelfs a, John F. Culling b, *, Mathieu Lavandier c
a
Welsh School of Architecture, Cardiff University, Bute Building, King Edward VIIAvenue, Cardiff CF10 3NB, UK School of Psychology, Cardiff University, Tower Building, Park Place, Cardiff, CF10 3AT, UK c Université de Lyon, Ecole Nationale des Travaux Publics de l’Etat, Département Génie Civil et Bâtiment (C.N.R.S.), Rue M. Audin, 69518 Vaulx-en-Velin Cedex, France
b
a r t i c l e i n f o
Article history: Received 27 May 2010 Received in revised form 1 December 2010 Accepted 6December 2010 Available online 13 December 2010
a b s t r a c t
Lavandier and Culling [Lavandier, M. and Culling, J. F. 2010. Prediction of binaural speech intelligibility against noise in rooms. J. Acoust. Soc. Am. 127, 387e399] demonstrated a method of predicting human speech reception thresholds for speech in combined noise and reverberation. An updated version of the model is presented,which is substantially more computationally efficient. The updated model makes similar predictions for the SRT data considered by Lavandier and Culling, which tested the model’s ability to predict effects of binaural unmasking and room colouration. In addition, we show here that the model accurately predicts the effects of headshadow and reproduces a range of data sets from the literature, includingsituations with multiple interfering sounds in anechoic conditions. Ó 2011 Elsevier B.V. All rights reserved.
1. Introduction Lavandier and Culling (2010) proposed a method of predicting Speech Reception Threshold (SRT) measurements for speech in combined noise and reverberation. Like previous models (vom Hovel, 1984; Zurek, 1993; Beutelmann and Brand, 2006; Beutelmann et al., 2010), themethod was based on a theory of binaural unmasking, combined with a model of better-ear listening. An updated version of the model has been developed which is substantially more computationally efficient. The method used by Lavandier and Culling (and also here) is illustrated schematically in Fig. 1. The first pathway calculates the expected binaural advantage due to binaural unmasking (Hirsh, 1948;Licklider, 1948) using Equalization-Cancellation theory (Durlach, 1963, 1972; Culling, 2007) to predict the Binaural Masking Level Difference (BMLD). The second path was designed to predict the benefits of better-ear listening. Lavandier and Culling did not include experiments to test whether the effects of headshadow on better-ear listening could be predicted, but some of their conditions werestrongly affected by room colouration, caused by reverberation, and they demonstrated that the model’s second path was essential to correctly predict these effects. Combined, the two paths should thus account for the two cues associated with spatial unmasking (Bronkhorst and Plomp, 1988), which is an established mechanism for the separation of competing sounds (Plomp, 1976; Hawley et al., 2004; Cullinget al., 2004), but the accurate
prediction of headshadow effects has not yet been demonstrated. Moreover, Lavandier and Culling only tested the model in situations with one interfering sound. The effectiveness of the model for stimuli that include headshadow and multiple, spatially distributed interferers is demonstrated here.
1.1. Specifics of the Lavandier and Culling model The model ofLavandier and Culling took as input speech-shaped noise, which had been convolved by Binaural Room Impulse Response (BRIR) recordings, to create reverberant speech-shapednoise interferers. Within each frequency channel, the waveforms were then processed through the two paths of the model independently. Different peripheral frequency channels were simulated by passing the waveforms through a...
Regístrate para leer el documento completo.