| Presentation | 2010-09-03 A Study of Reverberation Robust Speech Recognition Using Multi-Condition HMM Keisuke KATAYAMA, Yoshikazu MIYANAGA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Using a speech recognition system in reverberant and noisy environment, the recognition rate is seriously declined. In this paper, we study the reverberation and noise robust speech recognition system by multi-condition training. This method uses a clean speech data with reverberant speech data generated from clean speech data and artificial reverberant impulse response. First, we evaluate the recognition rate of the speech recognition experiment in each reverberant environments using the multi-condition HMM made of a clean speech data and one reverberant speech data. Based on this results, we show that the multi-condition training using plural reverberant speech data is effective in various reverberant environments. An observed signal is distorted because of noise influences in real environment. Therefore, we evaluate the multi-condition training that considers reverberant and noisy environment. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | speech recognition / multi-condition training / reverberation robust / noisy environment |
| Paper # | SIS2010-30 |
| Date of Issue |
| Conference Information | |
| Committee | SIS |
|---|---|
| Conference Date | 2010/8/26(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 | Smart Info-Media Systems (SIS) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A Study of Reverberation Robust Speech Recognition Using Multi-Condition HMM |
| Sub Title (in English) | |
| Keyword(1) | speech recognition |
| Keyword(2) | multi-condition training |
| Keyword(3) | reverberation robust |
| Keyword(4) | noisy environment |
| 1st Author's Name | Keisuke KATAYAMA |
| 1st Author's Affiliation | Graduate School of Information Science and Technology, Hokkaido University() |
| 2nd Author's Name | Yoshikazu MIYANAGA |
| 2nd Author's Affiliation | Graduate School of Information Science and Technology, Hokkaido University |
| Date | 2010-09-03 |
| Paper # | SIS2010-30 |
| Volume (vol) | vol.110 |
| Number (no) | 189 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |