Exploiting variable-width features in large vocabulary speech recognition.

M. Jones, Philip C. Woodland

Browse the full ICASSP paper archive.