Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Modelling the cocktail party: a binaural model for speech intelligibility in noise

Jelfs, Sam 2011. Modelling the cocktail party: a binaural model for speech intelligibility in noise. PhD Thesis, Cardiff University.
Item availability restricted.

[thumbnail of 2011JelfsSPhD.pdf]
Preview
PDF - Accepted Post-Print Version
Download (11MB) | Preview
[thumbnail of Jelfs.pdf] PDF - Supplemental Material
Restricted to Repository staff only

Download (2MB)

Abstract

We often listen to speech in an imperfect environment, with noise and reverberation; there will be voices around us, in complex acoustics. In this “cocktail-party” situation (Cherry 1953) listeners are helped by two binaural processes to segregate the desired voice from the competing noise: Better-Ear listening (BE) and Binaural Unmasking (BU). The aim of this thesis was to develop a model capable of efficiently predicting the benefits of BE and BU from Binaural Room Impulse Responses (BRIR). The developed model is a computationally efficient version of that created by Lavandier & Culling (2010) that predicts speech reception thresholds which include the benefit of binaural-unmasking, as explained by Equalization-Cancellation theory (Durlach 1963, 1972), and the benefit of better-ear listening, through Target-to-Interferer ratio analysis. The model accurately predicted a number of appropriate data sets from the literature that measure speech reception thresholds as a function of target and interferer source locations. Application of the model to a number of novel situations allowed environmental factors affecting intelligibility to be predicted and explored. In most situations, the effect of reverberation is to reduce the level of BE and BU, except when the listener is close to the interfering source, but this is when the benefits are needed the most. Depending on the spatial separation and source distances, the inclusion of multiple interferers again reduces the benefits in the majority of situations. To examine the benefits of head orientation a number of configurations were tested, whilst rotating the listener relative to the sound field. Benefits exceeding 12 dB can be achieved through modest rotations, particularly showing the benefits of BE. According to the model, the current literature on the benefits of bilateral cochlear implantation has underestimated that benefit by employing sub-optimal spatial configurations; using optimum orientations the model predicts an extra 6 dB of benefit being available to the listener. In a simulated restaurant situation, the model predicts that orientation of a table can affect the ability of a listener by up to 5 dB.

Item Type: Thesis (PhD)
Status: Unpublished
Schools: Architecture
Psychology
Uncontrolled Keywords: Noise; Reverberation; Acoustics; Binaural
Funders: EPSRC
Date of First Compliant Deposit: 30 March 2016
Last Modified: 19 Mar 2016 22:20
URI: https://orca.cardiff.ac.uk/id/eprint/12384

Citation Data

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics