2011 Fiscal Year Final Research Report
Regularization in a reproducing kernel Hilbert space for robust speech processing
| Project/Area Number |
22700193
|
|---|
| Research Category |
Grant-in-Aid for Young Scientists (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Perception information processing/Intelligent robotics
|
|---|
| Research Institution | National Institute of Information and Communications Technology |
|---|
Principal Investigator |
LU Xugang 独立行政法人情報通信研究機構, ユニバーサルコミュニケーション研究所音声コミュニケーション研究室, 専攻研究員 (20362022)
|
|---|
| Project Period (FY) |
2010 – 2011
|
| Keywords | 知能ロボット |
|---|
| Research Abstract |
In most current speech feature extraction methods, only low-order statistical information is extracted. In order to improve the robustness, we proposed a speech signal processing method in a reproducing kernel Hilbert space(RKHS). We first proved a theoretical analysis of the proposed framework, and showed a connection of the trade-off problem in machine learning and noise reduction.(From theoretical aspect, we proved that the trade-off problem in machine learning and noise reduction is essentially the same). Based on the theoretical analysis, we applied the framework for speech enhancement, and voice activity detection problems. Based on our application experiments, we showed that the proposed framework can be well used on speech processing and obtained improvement compared with several traditional noise reduction methods.
|
