Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

A New Tone Conversion Method for Mandarin by an Adaptive Linear Prediction Analysis

Congxiu Wang, Qihu Li, Guoying Zhao, Li Yin, Shuai Hao, Da Meng

Institute of Acoustics, Academia Sinica, Beijing, China

Conventional frame-based linear prediction coding(LPC) does not always de-convovle the speech signal into the vocal tract response and the voice source excitation signal exactly. In order to get better results, an adaptive linear prediction coding analysis(ALPC) method is proposed in this paper. The vocal tract responses obtained from ALPC analysis are used to synthesize a vowel syllable that has a different tone. During the synthesizing procedure, the voice source excitations are superseded by the improved LF-4 model, and the pitch-synchronized synthesizing method is used. The hearing experiments indicate that the synthesized speech by ALPC analysis, which has a different tone from the original one, has a high intelligibility and a better voice quality than that by the traditional LPC analysis method.


Full Paper

Bibliographic reference.  Wang, Congxiu / Li, Qihu / Zhao, Guoying / Yin, Li / Hao, Shuai / Meng, Da (2000): "A new tone conversion method for Mandarin by an adaptive linear prediction analysis", In ICSLP-2000, vol.2, 1088-1091.