Research Studentship/Fellowship on "THE ROLE OF FINE PHONETIC DETAIL IN NOISE"

[Guy Brown <g.brown@xxxxxxxxxxxxxx>]

PhD Research Studentship / Research Fellowship on
"THE ROLE OF FINE PHONETIC DETAIL IN NOISE"

University of Sheffield, UK, and Technical University of Cluj-Napoca,  
Romania

Two positions are available for research on the role of fine phonetic  
detail in noise, within the Marie Curie Research Training Network  
"Sound to Sense" (S2S) (www.sound2sense.eu). S2S involves  
collaborative speech research amongst 13 universities in 10  
countries. The position at the University of Sheffield, UK, is a PhD  
studentship with the Departments of Computer Science and Human  
Communication Sciences. The position at the Technical University of  
Cluj-Napoca, Romania, is a research fellowship, optionally including  
enrollment for a PhD. Both positions are 36 months in duration and  
are available immediately.

1) PhD research studentship on audio and visual cues to turn  
competition in overlapping talk, University of Sheffield, UK

There have been proposals in the conversation analysis literature,  
based on subjective listening, as to the various phonetic  
modifications that occur in different types of overlapping talk (i.e.  
when two or more people speak at the same time), including  
modification of pitch, loudness and timing. Current work at Sheffield  
University is developing techniques for objective measurement of  
these phonetic modifications in corpora of multi- talker recordings  
involving speakers varying in age, language and accent background.  
Depending on the interests and experience of the applicant, the  
proposed PhD studentship could develop this work in a number of  
directions, such as:

* Integrating visual information about speaker behaviour into the  
analysis, in order to understand the respective contributions of  
audio and visual information to the management of turn-taking and  
overlapping talk.

* Use of state-of-the-art motion tracking hardware to study the use  
of head-related gestures during overlapping and non-overlapping speech.

* The study of cross-language effects, in order to determine the role  
of language-specific cues in the management of overlap.

* Investigating the management of overlap in acoustic environments in  
which background noise (such as speech babble) is present.

* Investigating the management of overlap in talk between adults and  
young children.

* Investigating the management of overlap in talk involving speakers  
with communication difficulties (e.g. stammering / dysfluency;  
autistic spectrum disorders; developmental speech and language  
impairments)

The project will suit either a computer science graduate who is  
interested in learning about CA and phonetics; or a linguistics/  
speech sciences graduate who has experience of computer programming  
and is interested in learning about speech technology. The student  
will be based jointly at the Departments of Human Communication  
Sciences (www.sheffield.ac.uk/hcs) and Computer Science  
(www.sheffield.ac.uk/dcs).

2) Research fellowship on the role of fine phonetic detail in noise,  
Technical University of Cluj-Napoca, Romania

Why are listeners so much better than machines at recognising speech  
in everyday conditions? One hypothesis, supported by recent  
experiments, is that the speech signal contains fine phonetic detail  
(FPD) which is not noticeable in quiet conditions, but which is  
essential to decode speech in noise. What is not clear is which  
aspects of FPD are responsible for increased intelligibility, but two  
factors that may be implicated are grouping and glimpsing. Grouping  
refers to the combination of parts of the signal based on evidence  
that they have arisen from a common acoustic source (e.g.  
harmonicity). Glimpsing is based on the observation that listeners  
appear to identify speech based on partial "views" rather than  
complete signals. The aim of the fellowship/studentship is to  
determine how these factors might account for speech intelligibility  
and for automatic speech recognition in noise.

This project will suit a computer science or engineering graduate,  
who is interested in researching automatic speech recognition in  
noisy conditions and to learn more about auditory processing and  
phonetics.

The student will be based at the Technical University of Cluj-Napoca,  
Romania (www.utcluj.ro). Enrollment into a PhD program will be  
possible, and is encouraged. The university provides good technical  
research facilities as well as very affordable accommodation in an  
attractive setting.

Suitability

Applicants must have less than 4 years experience in postgraduate  
research (i.e. you will have earned your academic qualifications for  
graduate study less than 4 years before your S2S appointment starts).
The appointed persons will receive a standard EU stipend for the  
country of work, together with allowances for relocation and language  
lessons etc. Due to EU mobility requirements, you cannot be appointed  
to the country of your citizenship or to a country where you have  
lived for more than 12 of the last 36 months.

How to apply

Both appointments are available immediately and will remain open  
until filled. Informal enquiries can be made to Dr. Guy Brown  
(g.brown@xxxxxxxxxxxxxx) or Prof. Mircea Giurgiu  
(Mircea.Giurgiu@xxxxxxxxxxxxx).

Your formal application should comprise: a curriculum vitae including  
nationality and residence, 2 letters of reference, and statement of  
interests. Submit it electronically, preferably in pdf format, to Dr.  
Brown and Prof. Giurgiu, and also to the S2S Coordinator's assistant,  
Naomi Hilton:

Miss Naomi Hilton
soundsen@xxxxxxxxxxxxxxxx
Phone: +44 (0)1223 760822
Fax: +44 (0)1223 335053