Edinburgh Research Explorer

Dr Peter Bell

Reader in Speech Technology

Research Interests

My primary research focus is on automatic speech recognition (ASR).  I am particularly interested the following topics:
  • methods for cross-domain and cross-lingual adaptation
  • regularisation methods, particularly multi-task learning
  • algorithms for efficient alignment and search of audio data
  • lightly supervised and semi-supervised training
  • end-to-end and raw-waveform methods
  • ASR systems for minority or under-resourced languages
  • speech recognition and alignment of broadcast media

Qualifications

2010, PhD in Automatic Speech Recognition, University of Edinburgh.

2005, MPhil, Computer Speech, Text and Internet Technology, University of Cambridge.

2002, BA, Mathematics, University of Cambridge.

Biography

Peter Bell is a Reader in Speech Technology in the Centre for Speech Technology Research (CSTR) in the School of Informatics.  His research covers many different aspects of acoustic modelling for speech recognition.  After receiving MA and MPhil degrees from Cambridge University, he obtained his PhD from Edinburgh in 2010, supervised by Prof Simon King.  Since graduating, he has continued his research at CSTR, working primarily with Prof Steve Renals on several large projects funded by EPSRC, H2020 and others.  He is currently Co-Investigator on the IARPA-funded MATERIAL programme and the EPSRC-funded SpeechWave project.  From 2012-2019, Peter divided his time between the academic and commerical worlds as part-time speech scientist at Quorate Technology Ltd.  He carries out consultancy for a number of other companies. 

Research outputs

  1. Lattice-based lightly-supervised acoustic model training

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

  2. On Learning Interpretable CNNs with Parametric Modulated Kernel-based Filters

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

  3. Trainable Dynamic Subsampling for End-to-End Speech Recognition

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

View all (58) »

ID: 219421