Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

Word-Level F0 Range in Mandarin Chinese and its Application to Inserting Words into A Sentence

Dawei Xu (1), Hiroki Mori (2), Hideki Kasuya (2)

(1) Graduate school of Engineering; (2) Faculty of Engineering, Utsunomiya University, Japan

This paper considers an automatic voice response application in which a word utterance is inserted into a fixed carrier sentence. An important task here is to adjust the F0 contour of the inserted word according to the F0 context of the carrier sentence. Instead of generating the F0 contour on syllable basis, we employ an approach to adjust the F0 contour of the whole word. In this approach, two questions arise: (a) how to evaluate the F0 context and (b) how to adjust the F0 contour suitably for the context. We have found that the F0 contour of a word can be appropriately regulated in a tone-independent word-level F0 range (WF0R). After estimating the WF0Rs of the preceding and succeeding words, the WF0R of the inserted word is set at the mean of these WF0Rs. The F0 contour of the inserted word is then mapped to the WF0R taking into account the tone combination of the word. A perceptual evaluation experiment showed that the adjusted F0 was coordinated well with the context.


Full Paper

Bibliographic reference.  Xu, Dawei / Mori, Hiroki / Kasuya, Hideki (2000): "Word-level F0 range in Mandarin Chinese and its application to inserting words into a sentence", In ICSLP-2000, vol.3, 338-341.