Multi-level adaptive networks in tandem and hybrid ASR systems
- 1 May 2013
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 6975-6979
Abstract
In this paper we investigate the use of Multi-level adaptive networks (MLAN) to incorporate out-of-domain data when training large vocabulary speech recognition systems. In a set of experiments on multi-genre broadcast data and on TED lecture recordings we present results using of out-of-domain features in a hybrid DNN system and explore tandem systems using a variety of input acoustic features. Our experiments indicate using the MLAN approach in both hybrid and tandem systems results in consistent reductions in word error rate of 5-10% relative.Keywords
This publication has 28 references indexed in Scilit:
- Making Deep Belief Networks effective for large vocabulary continuous speech recognition2011
- Study of probabilistic and Bottle-Neck features in multilingual environment2011
- fMPE: Discriminatively Trained Features for Speech Recognition2006
- Cross-Domain and Cross-Language Portability of Acoustic Features Estimated by Multilayer Perceptrons2006
- A Fast Learning Algorithm for Deep Belief NetsNeural Computation, 2006
- On use of task independent training data in tandem feature extraction2004
- Recent improvements to the ABBOT large vocabulary CSR system2002
- Connectionist speech recognition of Broadcast NewsSpeech Communication, 2002
- State clustering in hidden Markov model-based continuous speech recognitionComputer Speech & Language, 1994
- Connectionist probability estimators in HMM speech recognitionIEEE Transactions on Speech and Audio Processing, 1994