The RWTH speech recognition system and spoken document retrieval

Author

Ney, H. ; Welling, L. ; Ortmanns, S. ; Beulen, K. ; Wessel, F.

Author_Institution

Lehrstuhl fur Inf., Tech. Hochschule Aachen, Germany

Volume

4

fYear

1998

fDate

31 Aug-4 Sep 1998

Firstpage

2022

Abstract

We present an overview of the RWTH Aachen large vocabulary continuous speech recognizer. The recognizer is based on continuous density hidden Markov models and a time-synchronous left-to-right beam search strategy. Experimental results on the ARPA Wall Street Journal (WSJ) corpus verify the effects of several system components, namely linear discriminant analysis, vocal tract normalization, pronunciation lexicon and cross-word triphones, on the recognition performance. Finally, the extension of the recognition system towards spoken document retrieval is discussed

Keywords

hidden Markov models; information retrieval; search engines; speech recognition; ARPA Wall Street Journal corpus; RWTH Aachen large vocabulary continuous speech recognizer; RWTH speech recognition system; continuous density hidden Markov models; cross-word triphones; linear discriminant analysis; pronunciation lexicon; recognition performance; spoken document retrieval; time-synchronous left-to-right beam search; vocal tract normalization; Acoustic beams; Cepstrum; Filters; Hidden Markov models; Linear discriminant analysis; Linear regression; Mel frequency cepstral coefficient; Speech recognition; Vectors; Vocabulary;

fLanguage

English

Publisher

ieee

Conference_Titel

Industrial Electronics Society, 1998. IECON '98. Proceedings of the 24th Annual Conference of the IEEE

Conference_Location

Aachen

Print_ISBN

0-7803-4503-7

Type

conf

DOI

10.1109/IECON.1998.724029

Filename

724029