2014 Fiscal Year Final Research Report
A method for identifying adult and child speaker based on auditory model for web system interface
| Project/Area Number |
25870438
|
|---|
| Research Category |
Grant-in-Aid for Young Scientists (B)
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Childhood science (childhood environment science)
Perceptual information processing
|
|---|
| Research Institution | Wakayama University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2013-04-01 – 2015-03-31
|
| Keywords | 音声情報処理 / ウェブシステム / こども / ディープニューラルネットワーク / 言語情報 / インタフェースデザイン |
|---|
| Outline of Final Research Achievements |
In this study, we have developed web-based speech interface system to identify adult and child speakers as follows: (1) Auditory features were integrated into the identification algorithms to improve accuracies. We have introduced deep learning neural networks (DNNs). (2) Interactive strategy method of determining language information were investigated. Co-occurrence information of vocabulary included in the newspaper articles and Bag-of-Words features were tested as linguistic features. (3) We have improved the proposed system as web service. The system is able to run on the Android terminal in addition to the PC terminal. We also studied about the design of the voice-enabled web user interface on the basis of the experimental results.
|
| Free Research Field |
音情報処理
|
