Show simple item record

IEEE Transactions on Audio, Speech and Language Processing

dc.contributor.authorGarau, Giuliaen
dc.contributor.authorRenals, Steveen
dc.date.accessioned2010-10-04T14:09:32Z
dc.date.available2010-10-04T14:09:32Z
dc.date.issued2008
dc.identifier.urihttp://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4443886
dc.identifier.urihttp://hdl.handle.net/1842/3820
dc.description.abstractIn this paper we investigate the combination of complementary acoustic feature streams in large vocabulary continuous speech recognition (LVCSR). We have explored the use of acoustic features obtained using a pitch-synchronous analysis, STRAIGHT, in combination with conventional features such as mel frequency cepstral coefficients. Pitch-synchronous acoustic features are of particular interest when used with vocal tract length normalisation (VTLN) which is known to be affected by the fundamental frequency. We have combined these spectral representations directly at the acoustic feature level using heteroscedastic linear discriminant analysis (HLDA) and at the system level using ROVER. We evaluated this approach on three LVCSR tasks: dictated newspaper text (WSJCAM0), conversational telephone speech (CTS), and multiparty meeting transcription. The CTS and meeting transcription experiments were both evaluated using standard NIST test sets and evaluation protocols. Our results indicate that combining conventional and pitch-synchronous acoustic feature sets using HLDA results in a consistent, significant decrease in word error rate across all three tasks. Combining at the system level using ROVER resulted in a further significant decrease in word error rate.en
dc.titleCombining Spectral Representations for Large Vocabulary Continuous Speech Recognitionen
dc.typeArticleen
dc.identifier.doi10.1109/TASL.2008.916519
rps.issue3en
rps.volume16en
rps.titleIEEE Transactions on Audio, Speech and Language Processingen
dc.extent.pageNumbers508--518en
dc.date.updated2010-10-04T14:09:33Z


Files in this item

This item appears in the following Collection(s)

Show simple item record