Distant-talking speech recognition based on spectral subtraction by multi-channel least mean square approach
| Project/Area Number |
22700169
|
|---|
| Research Category |
Grant-in-Aid for Young Scientists (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Perception information processing/Intelligent robotics
|
|---|
| Research Institution | Nagaoka University of Technology (2012)
Shizuoka University (2010-2011) |
|---|
Principal Investigator |
WANG Longbiao 長岡技術科学大学, 産学融合トップランナー養成センター, 産学融合特任准教授 (30510458)
|
|---|
| Project Period (FY) |
2010 – 2012
|
| Project Status |
Completed (Fiscal Year 2012)
|
| Budget Amount *help |
¥3,900,000 (Direct Cost: ¥3,000,000、Indirect Cost: ¥900,000)
Fiscal Year 2012: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2011: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2010: ¥1,690,000 (Direct Cost: ¥1,300,000、Indirect Cost: ¥390,000)
|
|---|
| Keywords | 一般化スペクトルサブトラクション / ハンズフリー音声認 / missing feature theory / マルチチャンネルLMS / ブラインド残響除去 / ハンズフリー音声認識 / 音源分離 / 独立成分分析 |
|---|
| Research Abstract |
We proposed a blind dereverberation method based on spectral subtraction using a multi-channel least mean square algorithm (MCLMS). This method was evaluated in a simulated and real noisy reverberant environment with stationary noise. In this study, we also evaluate this method in a noisy reverberant environment with non-stationary noise like music. After suppressing the music, using a blind source separation based on Efficient FastICA (independent component analysis) algorithm, spectral subtraction based dereverberation method is employed to reduce late reverberation. The proposed method achieves an average relative word error reduction rate of 41.9% and 7.9% compared to baseline method and the state-of-art multi-step linear prediction (MSLP) based dreverberation in a real environment, respectively.
|
Report
(4 results)
Research Products
(48 results)
