PHD POSITION IN COMPUTATIONAL MODELLING OF HEARING IMPAIRMENT

[Guy Brown <g.brown@xxxxxxxxxxxxxx>]

RNID RESEARCH STUDENTSHIP - COMPUTATIONAL MODELLING OF SPEECH 
PERCEPTION IN HEARING IMPAIRED LISTENERS BASED ON MISSING FEATURE 
THEORY

A research studentship in computational modelling of speech perception 
in hearing impaired listeners is available at the Department of 
Computer Science, University of Sheffield, UK, funded by the RNID.

The studentship provides a stipend (approx. £15,000 p.a.), 
conference/training allowance (approx. £1,000 p.a.) and UK/EU tuition 
fees for PhD study. Applicants from outside the UK/EU will be 
considered if they are able to cover the difference between the UK/EU 
and international tuition fee. The studentship will run for three years 
and will ideally start in October 2007.

The aim of the studentship is to develop a computational model of 
speech recognition in hearing impaired listeners, based on a 'missing 
feature' approach. Missing feature techniques have been used to improve 
the robustness of automatic speech recognition in noisy conditions by 
exploiting information about the relative levels of speech and noise in 
each time-frequency region of the acoustic signal. The speech 
recogniser uses different strategies to deal with 'reliable' regions 
(which are dominated by speech) and 'unreliable' regions (which are 
dominated by noise). In this studentship, a similar approach will used 
to model speech recognition in hearing impaired listeners, using the 
characteristics of a particular listener's hearing impairment (e.g., 
their susceptibility to simultaneous and forward masking) to determine 
the 'reliable' and 'unreliable' time-frequency regions. More 
specifically, the aims of the studentship are:

* To develop a missing feature model of hearing impairment;
* To investigate the utility of the missing feature model in 
characterising the hearing impairment of particular subjects, and in 
predicting the confusions that listeners will make when recognising 
speech in a background of noise;
* To validate the model by comparing it against psychometric data from 
hearing impaired subjects, and by using it to inform the design of 
'proof of concept' hearing aid algorithms.

The studentship is part of an ongoing collaboration with Prof. Ray 
Meddis of the Department of Psychology, University of Essex. The 
successful applicant will be based at Sheffield, but will spend some 
time at Essex in order to attend progress meetings and receive training 
in psychometric testing techniques. Applicants should have a strong 
background in computer science, psychology or a related discipline.

Applicants should send a full curriculum vitae, including the names and 
addresses of three referees, to:

Dr. Guy Brown
Speech and Hearing Research Group
Department of Computer Science
Regent Court, 211 Portobello Street
Sheffield S1 4DP
United Kingdom

The deadline for applications is 1st May 2007. Informal enquiries can 
be made to Dr. Brown by email (g.brown@xxxxxxxxxxxxxx) or phone (+44 
114 222 1821).

See http://www.dcs.shef.ac.uk/spandh/ for further information.