Sciweavers

ICASSP
2011
IEEE

A multi-stream ASR framework for BLSTM modeling of conversational speech

12 years 7 months ago
A multi-stream ASR framework for BLSTM modeling of conversational speech
We propose a novel multi-stream framework for continuous conversational speech recognition which employs bidirectional Long Short-Term Memory (BLSTM) networks for phoneme prediction. The BLSTM architecture allows recurrent neural nets to model longrange context, which led to improved ASR performance when combined with conventional triphone modeling in a Tandem system. In this paper, we extend the principle of joint BLSTM and triphone modeling to a multi-stream system which uses MFCC features and BLSTM predictions as observations originating from two independent data streams. Using the COSINE database, we show that this technique prevails over a recently proposed single-stream Tandem system as well as over a conventional HMM recognizer.
Martin Wöllmer, Florian Eyben, Björn Sch
Added 20 Aug 2011
Updated 20 Aug 2011
Type Journal
Year 2011
Where ICASSP
Authors Martin Wöllmer, Florian Eyben, Björn Schuller, Gerhard Rigoll
Comments (0)