To work on robust methods for direct recognition from the raw waveforms - to achieve robustness against environmental effects. Work should focus on telephony, but further multichannel input is of interest. Funding for 3 years, in the context of the Voicebase Centre for Speech and Language Technologies.