| Paper: | SP-L5.3 | ||
| Session: | Pitch and Tone Based Speech Analysis | ||
| Time: | Thursday, May 20, 10:10 - 10:30 | ||
| Presentation: | Lecture | ||
| Topic: | Speech Processing: Speech Analysis | ||
| Title: | TONE RECOGNITION WITH FRACTIONIZED MODELS AND OUTLINED FEATURES | ||
| Authors: | Ye Tian; Microsoft Research Asia | ||
| Jian-Lai Zhou; Microsoft Research Asia | |||
| Min Chu; Microsoft Research Asia | |||
| Eric Chang; Microsoft Research Asia | |||
| Abstract: | In this paper, different feature extraction and tone modeling schemes are investigated on both speaker-dependent and speaker-independent continuous speech database. Tone recognition features can be classified as detailed features which use the entire F0 curve, and outlined features which capture the main structure of the F0 curve. Tone models of different size, ranging from very simple one-tone-one-model tone models to complex phoneme-dependent tone models, have different ability to characterize tone. Our experiments show two conclusions: 1) the detailed information of the F0 curve is not necessary for tone recognition. The outlined features can not only reduce the number of parameters, but also improve the accuracy of tone recognition. The subsection average F0 and ?F0, proposed in this paper, is shown to be effective outlined features. 2) The one-tone-one-model scheme is not sufficient. Building phoneme-dependent tone models can highly improve the tone recognition accuracy, especially for speaker-independent data. Thus we suggest use fractionized models trained with outlined features for tone recognition. | ||
| Back | |||
Home -||-
Organizing Committee -||-
Technical Committee -||-
Technical Program -||-
Plenaries
Paper Submission -||-
Special Sessions -||-
ITT -||-
Paper Review -||-
Exhibits -||-
Tutorials
Information -||-
Registration -||-
Travel Insurance -||-
Housing -||-
Workshops