Audio Technology

A common cause of Aural diversity is hearing loss, with 12 million people in the UK having hearing loss in both ears. The main treatment for this is hearing aids, but the performance of these still needs improvement for speech intelligibility in noisy places. Machine learning has great potential to improve speech in noise. This PhD is designed to build on the The Clarity Project (https://claritychallenge.org/) which has been running a series of machine learning challenges to improve the processing of speech in noise on hearing aids.

Your PhD will involve developing and running new challenges working with members of the Clarity team. Opportunities for novel research will arise from this. To give two examples: (1) You might develop a better hearing loss model for machine learning to be part of a software baseline, or (2) You might develop new listening test methods that allows more ecologically-valid assessment of the audio. Applicants for this PhD need to be competent coders. Preferably with experience of using Python, Git and machine learning frameworks

Supervisors: Trevor Cox and Ian Drumm

There’s a proliferation of virtual display technologies for entertainment, training and marketing, among other applications. Typically, content in these platforms does not comply with needs for those with diverse aural perception modes. This project will research how these technologies may be made more inclusive for those with hearing difference. The project will entail creation of environments in VR/AR platforms and the design of subjective testing to understand response to inclusive strategies. Skill in programming, production of immersive audio environments and design and deployment of subjective testing are desirable.

Supervisors: Bruno Fazenda and Chris Hughes

Response to synthetic voices is an understudied research topic, and as more and more AI assistants are being developed to help with a myriad of societal issues, it is important to understand how these voices are processed. When judging synthetic voices, what do listeners think is behind the voice? How are peoples’ judgements influenced by stereotypes that they use with human voices? What are the individual differences in how synthetic voices are tolerated and embraced? What are the implications of this? For example, how do different individuals process information from digital assistants? In this project you will develop experimental methods to answer the key questions posed. Qualitative methods can also be employed to investigate the subjective user experience.

Supervisors: Sam Gregory and Trevor Cox

Some aurally divergent individuals can find many environmental sounds as being disturbing, confusing or even painful to hear. Recent work in AI has improved the capability of sound source separation, sound (e.g., speech) enhancement and reduction (e.g., background noise). This project would investigate deep learning techniques for sound identification and separation with the aim of facilitating real time rebalancing of environmental sounds based on individual requirements or needs.

Supervisors: Ben Shirley and Chris Hughes

Anecdotal reports note the stress-inducing sonic environments of ICU wards, particularly under pandemic conditions. Much information vital to healthcare is conveyed through sound, but staff report difficulties in streaming the audio content appropriately, and in localising monitor sounds, creating a ‘wall of noise’ effect. Existing research suggests reduced auditory streaming to be affected by the presence of non-neurotypical conditions; also evidenced is the ability of trained musicians to stream and localise sounds more effectively. Could training healthcare staff to listen more effectively using musical training assist them in identifying and localising information-bearing sounds in the clinical environment?

Supervisors: Alan Williams and Phil Brissenden

Auracast is a new broadcast audio technology which promises to revolutionise access to audio in private and public spaces. The technical specifications define how devices communicate, but not how they should be used. Leaning on research in acoustics, technology and psychology this project offers an opportunity to explore the perceived benefits of assistive listening and what system designers, installers and managers need to do to ensure that these benefits are delivered for those with hearing diversity.

Supervisors: Trevor Cox and Adam Galpin