PhD Studentship: Generating realistic, expressive speech animation (THEOBALD_U15EP)

University of East Anglia - School of Computing Science

Start Date: October 2015.

Supervisor: Dr Barry-John Theobald, b.theobald<στο>uea.ac.uk

The Project:

Creating believable facial animation is a difficult task. Humans read important cues from subtle motion on the face, and so any facial movements displayed on an animated character must match with viewer expectation given the speech content, the expressive style and the context. Consequently, visual effects studios resort to time-consuming and expensive motion-capture or handcrafted animation by skilled artists to create content for computer games and animated movies and TV shows. It is common for animators to spend several hours creating only a few seconds of expressive dialogue.

There has been interest over the last decade or so in methods for automating the creation of expressive animated speech.  State of the art approaches typically handle only a small number of expressive speaking styles and they require a large amount of training data for each style.  We recently developed a new approach that learns a single model for a range of expressive styles from only a few short sentences.  A small set of prototypical facial expressions (e.g. happy, angry, etc.) are used to learn the range of motion required for all expressive styles, and then more subtle expressions that are difficult to capture (e.g. interest, confusion, etc.) are generated without requiring an explicit model for each.

This PhD project will build upon our existing work to investigate how conversational expressions can be coded and modelled without requiring explicit emotional labels.  In particular, automated methods for clustering the expressions and relating the underlying clusters back to the speech content will be investigated so that these models are useful to animators in a practical sense.

Entry Requirements:

Applicants should have a minimum of a 2.1 degree in computing, engineering or a mathematics related subject.

Funding:

This PhD project is in a Faculty of Science competition for EPSRC funded studentships. These studentships are funded for 3 years* and comprise home/EU fees, an annual stipend of £13,863 and £1000 per annum to support research training

* Funding for EPSRC studentships is available to successful candidates who meet the UK Research Council eligibility criteria including the 3-year UK residency requirements. These requirements are detailed in the EPSRC eligibility guide which can be found athttp://www.epsrc.ac.uk/skills/students/help/Pages/eligibility.aspx. In most cases UK and EU nationals who have been ordinarily resident in the UK for 3 years prior to the start of the course are eligible for a full-award. Other EU nationals may qualify for a fees only award, and the Faculty has some funding available to support exceptional EU candidates in order for them to receive a full award.

Interviews will be held on: 6 and 9 March 2015

To discuss the application process or particular projects, please contact the: Admissions Office, email:pgr.enquiries.admiss<στο>uea.ac.uk or telephone +44 (0)1603 591709. 

Apply