Output details
11 - Computer Science and Informatics
Queen's University Belfast
Combining missing-feature theory, speech enhancement, and speaker-dependent/-independent modeling for speech separation
<22>This paper is in a CSL special issue, containing a selection of invited submissions on the first international challenge for speech separation. The challenge is about the separation of two mixed speech utterances spoken by the same or two different speakers, with an extremely low signal-to-noise ratio (as low as -9 dB) and recorded using a single microphone. This paper described the system developed jointly by the Queen’s University Belfast and the Massachusetts Institute of Technology, USA, for this competition. The system was awarded first place in the UK and third internationally.