| Presentation | 2000/12/14 Extended Kalman Particle filters applied to model-based noise compensation for noisy speech recognition Kaisheng Yao, Tomoko Matsui, Satoshi Nakamura, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | We suggest viewing noisy speech recognition based on Jump Markov State Space model. In this model, noise parameters and state sequences are hidden and estimated by a computational Bayesian approach for parameter estimation. Particularly, the Monte-Carlo particle filters were adopted to estimate time-varying additive noise parameter for model-based noise compensation. Each particle corresponds to a certain state space of noise. The particles randomly transit to new state spaces of noise according to the transition probability given by acoustic models and language models for speech recognition. Higher likelihood particles generate larger number of new particles with newly evolved state space, whereas the lower likelihood particles may be stopped by a selection step. The state space after a particular transition was evolved using an extended Kalman filter. Likelihood of each state space contributes to Minimum Mean Square Error (MMSE) estimation of the noise parameter from all the particles. Primary experiments on N-Best rescoring are shown in this paper. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Speech recognition / Noise compensation / State space model / Kalman filter / Monte-Carlo method / Particle filter |
| Paper # | NLC2000-31,SP2000-79 |
| Date of Issue |
| Conference Information | |
| Committee | NLC |
|---|---|
| Conference Date | 2000/12/14(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Natural Language Understanding and Models of Communication (NLC) |
|---|---|
| Language | ENG |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Extended Kalman Particle filters applied to model-based noise compensation for noisy speech recognition |
| Sub Title (in English) | |
| Keyword(1) | Speech recognition |
| Keyword(2) | Noise compensation |
| Keyword(3) | State space model |
| Keyword(4) | Kalman filter |
| Keyword(5) | Monte-Carlo method |
| Keyword(6) | Particle filter |
| 1st Author's Name | Kaisheng Yao |
| 1st Author's Affiliation | ATR Spoken Language Translation Research Laboratories() |
| 2nd Author's Name | Tomoko Matsui |
| 2nd Author's Affiliation | ATR Spoken Language Translation Research Laboratories |
| 3rd Author's Name | Satoshi Nakamura |
| 3rd Author's Affiliation | ATR Spoken Language Translation Research Laboratories |
| Date | 2000/12/14 |
| Paper # | NLC2000-31,SP2000-79 |
| Volume (vol) | vol.100 |
| Number (no) | 520 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |