[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
PhD Research Studentship in Microphone Array based Acoustic Event Analysis
We have a PhD Research Studentship in microphone array based acoustic event analysis, at the University of Edinburgh. This studentship, to start in September 2014, will be jointly supervised by Steve Renals and James Hopgood.
The big picture of this research project is to identify and characterise the sources that make up an acoustic scene captured using a microphone array. Acoustic scene analysis and distant speech recognition has applications in surveillance, both civilian and homeland security, assisted living environments, hands-free distant human-computer interfaces, forensics, and free-form diarization. This work extends our recent work of microphone array based speech recognition and speaker diarization. In particular, previous work has focused entirely on speech sources, focusing on those segments that have been identified as containing speech. However if we wish to fully understand a scene from the acoustics then it is important to model all acoustic events, not just speech.
Our approach to this problem will be to use graph transformer networks, in which different modules are all trained using a single objective function and gradient descent based training. This is an interesting extension to our ongoing work using modular deep neural networks for distant speech recognition. In particular we are interested to explore using the graph transformer network approach to include the signal processing components which are currently either not learned automatically, or optimised separately.
This PhD will be funded by an EPSRC studentship, for this reason only UK applicants, or EU applicants who have lived in the UK for more than three years, are eligible.
Application deadline: 7 February 2014.
Please see http://www.cstr.ed.ac.uk/opportunities/ for more details.
University of Edinburgh
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.