I am a senior research fellow at The Centre for Speech Technology
Research. I conduct research into machine learning techniques for
speech processing, in particular acoustic modelling for automatic
speech recognition (ASR). My goals are to improve the accuracy and
portability of ASR systems, and to develop search engines for audio
material.
The themes of my work to date are the use of knowledge about speech
production in models for automatic speech recognition, investigation
of subword units other than phones, including factored representations
such as acoustic/articulatory features, and improving robustness and
portability of ASR systems through transfer learning. I also conduct
research on other spoken language technologies, including speech-based
information retrieval and dialogue act recognition.
My long-term strategy is to develop a consistent machine-learning
framework in which all components of spoken language processing
systems can be jointly optimised with respect to a single objective
function, which would be dependent on the end use. For example for
automatic speech recognition, this might be minimum word error rate.
More information on my work, including publications and PhD thesis,
can be found here
or on the main CSTR publications
page.