FHMM for Robust Speech Recognition in Home Environment
Download:
by Agnieszka Betkowska, Koichi Shinoda, Sadaoki Furui
http://www.furui.cs.titech.ac.jp/english/publication/../../publication/2006/3-p009.pdf
Add To MetaCart
Abstract:
In this paper, we focus on the problem of speech recognition in the presence of nonstationary sudden noise that appears rapidly and lasts for a short period of time. As a model compensation for this task, we investigated the use of Factorial Hidden Markov Model (FHMM) architecture built from clean speech Hidden Markov Models (HMMs) and sudden noise HMM. As this architecture is defined only for static features of the observation vector, we extended it also for dynamic features. The experiments confirmed that the proposed method improves the clean speech HMMs, particularly in noisy conditions. 1.
Citations
| 384 | Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences – Davis, Mermelstein - 1980 |
| 118 | Hidden Markov model decomposition of speech and noise – Varga, Moore - 1990 |
| 106 | Speaker-independent isolated word recognition using dynamic features of speech spectrum – Furui - 1986 |
| 104 | Spoken Language Processing: A Guide to Theory, Algorithm and System Development – Huang, Acero, et al. - 2001 |
| 95 | Robust automatic speech recognition with missing and unreliable acoustic data – Cooke, Green, et al. - 2001 |
| 21 | Speech Recognition Using Noise-Adaptive Prototypes – Nadas, Nahamoo, et al. - 1989 |
| 7 | Factorial HMMs for acoustic modeling,” ICASSP – Logan, Moreno - 1998 |
| 6 | A factorial HMM approach to simultaneous recognition of isolated digits spoken by multiple talkers on one audio channel – Deoras, Hasegawa-Johnson - 2004 |
| 2 | A Speech Recognition Interface for Robots using Notification of III-Suited Conditions – Iwasawa, Ohnaka, et al. - 2002 |
