Understanding British Accents in Noise

Accent differences among speakers and listeners can interfere with the ability of individuals to understand each other under noisy conditions. For example, listeners are best at understanding speech that matches either their own accent or standard accents (e.g., southern British accents that are frequently used in the media), but other accents can rapidly deteriorate in noise. These accent effects can be large, similar to the effect of having a mild hearing loss. Noise and accents can be particularly problematic for children, older adults, and individuals with hearing impairments. These accent effects in noise thus have practical importance for real-world speech communication, but the investigation of these effects also has broader scientific implications, allowing us to address fundamental issues of speech plasticity and the architecture of human speech processing.

The overall aim of this research is to understand why and how accent differences among British English speakers and listeners can make speech recognition difficult in noisy conditions. We will test whether (i) listeners are multidialectal (i.e., able to understand many different accents, not only ones that are similar to their own), and (ii) whether the difficulties with accents and noise come from early stages of speech processing in the brain, or at later processing stages associated with the recognition of words. Our experiments will involve testing individuals with a wide range of British English accents. We will test their ability to recognize speech spoken in different accents and mixed with noise, acoustically analyze their own speech in terms of accent, and use neurophysiological measures (Electroencephalography; EEG) to assess different types of speech processing. This interdisciplinary collection of measures and techniques will be used to address questions that are relevant to sociophonetics (e.g., why certain accents become standards, and the impacts of multidialectal experience), speech science (e.g., which factors explain speech intelligibility) and psychology (e.g., how speech is processed in the brain, and the perceptual learning abilities of adults), and do so in a way that will have practical relevance to understanding how people communicate in the UK.

Audiological testing. This project will investigate the intelligibility of standard accents by a range of UK accent groups under challenging conditions. This is significant from an audiological perspective, because standard accents predominate among the speech materials that are used for hearing testing. Many of the standard UK speech test materials are distributed by the MRC Institute of Hearing Research, with some of these materials being produced in our department at UCL by Andrew Faulkner. If we find that non-standard accents are required by some listener groups for accurate assessment (i.e., standard accents would underestimate their hearing abilities), the sentence recordings generated in our proposed work can be repackaged to create a new battery of clinical test materials, although additional small-grant support (e.g., RNID, Deafness Research UK) will likely be required to validate the tests. In addition, the findings from this project will be disseminated at a meeting of the British Society of Audiology, with an audience of clinicians and hearing researchers.

Speech Synthesis. Telecommunications and speech synthesis companies also tend to use standard accents for their applications. As in the audiology impact described above, our work will produce data on which of these accents are likely to be most intelligible to different listener groups. We will make our work known to leading telecommunication and speech technology companies, by disseminating our results at workshops that we will hold as part of the new Marie Curie Initial Training Network INSPIRE (Investigating Speech Processing In Realistic Environments, 2012-2016). The network will support postgraduate research on speech processing under realistic conditions, and our network includes many industrial partners (e.g., Philips Research Laboratories, Nokia, Nuance, Toshiba, Cochlear) as well as academic partners who are engaged in commercial speech applications (e.g., Edinburgh). The network workshops will thus be ideal for disseminating our results, and allow us to discuss commercial exploitation of our findings.

General public and society. Accents and speech are topics that are of general popular interest. We have been active in public dissemination, with interviews given for radio (BBC Radio Four, BBC World Service, BBC Scotland, BBC London), UK national newspapers (Guardian, Observer, Telegraph), international news media (Die Zeit, Germany), and the popular-science press (Scientific American). We have recently presented our work at a fully booked public science evening at the British Library (http://www.bl.uk/reshelp/experthelp/science/science@blevents/previousevents/auditoryillusions/yourbrainandsound.html), and at the 2011 Royal Society Summer Exhibition (http://royalsociety.org/summer-science/2011/noisy-world/). We have written an iPhone App for second-language vowel training based on one of our previous ESRC-funded projects (RES-000-23-0838); it has had more than 3,000 downloads thus far (http://itunes.apple.com/gb/app/british-english-vowel-training/id401156507?mt=8).

Any newsworthy findings from the present project will be disseminated to the media via a UCL press release and accompanying interviews. We will also produce an iPhone App to evaluate the accent of the listener (e.g., Which British English accent do you have?) or demonstrate the effects of noise on speech recognition, either of which could serve as an opportunity for press contacts. Given our experience with writing successful iPhone Apps, this will not require significant expenditure beyond the research funding that is requested (e.g., PI effort and software updates). Our next scheduled public science event is in summer 2016 (after this grant is completed), but public engagement is a regular part of our academic practice, and we will promote the findings of this project at these types of events as they arise.