Research
I am interested in spoken language understanding for both humans and machines. At the moment my research interests include:
- Hearing impairment and cochlear implant processing
- Computational auditory scene analysis
- Noise-robust automatic speech recognition
- Speech perception in noise
CHiME
My latest research was funded by the EPSRC project CHiME: Computational Hearing in Multisource Environments, which aims to develop a computational model of human hearing ability in natural listening environments where a large and variable number of sound sources are present. The model will be based on a unified statistical framework to link deterministic signal processing that mimics the auditory periphery to probabilistic model-driven processing which embodies knowledge of sound sources.
We organised the 2011 PASCAL CHiME Speech Separation and Recognition Challenge, hosted at the ISCA Workshop on Machine Listening in Multisource Environments, Florence, 1st September 2011.
AMORPH
Before CHiME I worked on the EPSRC-funded interdisciplinary research project AMORPH: Amorphous Computing, Random Graphs and Complex Biological Systems, during which I developed computational models using graph and network theory within a variety of biological and technological domains.
Ph.D. Research
My doctoral study (completed in 2008) focused on investigating a human-inspired approach to automatic speech recognition in noisy environments. I developed novel techniques for identifying acoustic events in sound mixtures and using this information to mediate sound organisation by a combination of bottom-up and top-down processing (see examples). The work has demonstrated remarkably robust speech recognition results in a competitive international evaluation: the Speech Separation Challenge. A copy of my Ph.D. thesis can be found here.
My Ph.D. work has contributed to the EC HOARSE (EU Framework 5) research training network, the EU project POP (EU Framework 6) and the SPECS project (funded by the Department of Health HTD programme).
I have been a visiting research scientist at the Department of Electrical Engineering, University of Washington, Seattle, during which I worked with Jeff Bilmes on the Multidisciplinary University Research Initiative project `Human-like Speech Processing'. This work investigated the application of graphical models to the problem of robust conversational speech recognition. See demonstrations and publication of the research.