| Presentation | 2006/12/14 Tone Recognition of Continuous Mandarin Speech Based on Tone Nucleus Model and Neural Network Xiao-Dong Wang, Keikichi Hirose, Jin-Song Zhang, Nobuaki Minematsu, Chen-Yu Chiang, Yih-Ru Wang, Yuan-Fu Liao, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Automatic tone recognition of Mandarin continuous speech is still far from a resolved task due to complex feature variations for the five tones. In this paper we propose to integrate two independent techniques together to challenge the task: The first one is the tone nucleus modeling that specifies a critical segment in the F_0 contour of a syllable to avoid influences from transitional F_0 loci. Here, a method of automatically detecting tone nucleus is crucial. We have developed such a method. The second one is to adopt MLP (Multi-layer Perceptron) as tone recognizer to accept heterogeneous features correlated to tones but difficult to be used in a HMM (Hidden Markov Model) tone recognizer. Experiments showed: When the two individual techniques were adopted independently, HMM-based tone nucleus modeling got 14.5% tone recognition error, and MLP recognizer got 14.1%. On the other hand, the integrated approach achieved a relative error reduction by about 10%, and got 12.8% tone recognition error. The improved performances indicate the proposal is at a right direction. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Mandarin / Tone Recognition / Tone Nucleus model / Automatic Tone Nucleus Extraction / MLP |
| Paper # | NLC2006-47,SP2006-103 |
| Date of Issue |
| Conference Information | |
| Committee | SP |
|---|---|
| Conference Date | 2006/12/14(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 | Speech (SP) |
|---|---|
| Language | ENG |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Tone Recognition of Continuous Mandarin Speech Based on Tone Nucleus Model and Neural Network |
| Sub Title (in English) | |
| Keyword(1) | Mandarin |
| Keyword(2) | Tone Recognition |
| Keyword(3) | Tone Nucleus model |
| Keyword(4) | Automatic Tone Nucleus Extraction |
| Keyword(5) | MLP |
| 1st Author's Name | Xiao-Dong Wang |
| 1st Author's Affiliation | The University of Tokyo() |
| 2nd Author's Name | Keikichi Hirose |
| 2nd Author's Affiliation | The University of Tokyo |
| 3rd Author's Name | Jin-Song Zhang |
| 3rd Author's Affiliation | NiCT |
| 4th Author's Name | Nobuaki Minematsu |
| 4th Author's Affiliation | The University of Tokyo |
| 5th Author's Name | Chen-Yu Chiang |
| 5th Author's Affiliation | National Chiao Tung University |
| 6th Author's Name | Yih-Ru Wang |
| 6th Author's Affiliation | National Chiao Tung University |
| 7th Author's Name | Yuan-Fu Liao |
| 7th Author's Affiliation | National Taipei University of Technology |
| Date | 2006/12/14 |
| Paper # | NLC2006-47,SP2006-103 |
| Volume (vol) | vol.106 |
| Number (no) | 443 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |