Recurrent Deep Stacking Networks for Speech Recognition

TL;DR

This paper introduces Recurrent Deep Stacking Networks (RDSNs) and a more efficient Bi-Pass Stacking Network (BPSN) for speech recognition, enhancing acoustic models with phoneme-level information to improve performance.

Contribution

The paper proposes RDSN and BPSN models that incorporate phoneme-level data into acoustic models, offering improved speech recognition accuracy.

Findings

RDSN and BPSN outperform conventional DNNs in speech recognition tasks.

BPSN provides a comparable performance to RDSN with higher efficiency.

Both models significantly improve robustness in ASR.

Abstract

This paper presented our work on applying Recurrent Deep Stacking Networks (RDSNs) to Robust Automatic Speech Recognition (ASR) tasks. In the paper, we also proposed a more efficient yet comparable substitute to RDSN, Bi- Pass Stacking Network (BPSN). The main idea of these two models is to add phoneme-level information into acoustic models, transforming an acoustic model to the combination of an acoustic model and a phoneme-level N-gram model. Experiments showed that RDSN and BPsn can substantially improve the performances over conventional DNNs.