Publications -> Conference Papers

Average Modeling Approach to Voice Conversion with Non-Parallel Data

Authors: X. Tian, J. Wang, H. Xu, E. S. Chng, and H. Li
Title: Average Modeling Approach to Voice Conversion with Non-Parallel Data
Abstract: Voice conversion techniques typically require source-target parallel speech data for model training. Such parallel data may not be available always in practice. This paper presents a nonparallel data approach, that we call average modeling approach. The proposed approach makes use of a multi-speaker average model that maps speaker-independent linguistic features to speaker dependent acoustic features. In particular, we present two practical implementations, 1) to adapt the average model towards target speaker with a small amount of target data, 2) to present speaker identity as an additional input to the average model to generate target speech. As the linguistic feature and the acoustic feature can be extracted from the same utterance, the proposed approach doesn’t require parallel data in either average model training or adaptation. We report the experiments on the voice conversion challenge 2018 (VCC2018) database that validate the effectiveness of the proposed method.
Keywords: Voice conversion; Non-parallel data; Average modeling approach (AMA)
Conference Name: Odyssey 2018
Location: Les Sables d'Olonne, France
Publisher: ISCA
Year: 2018
Accepted PDF File: Average_Modeling_Approach_to_Voice_Conversion_with_Non-Parallel_Data_accepted.pdf
Permanent Link:
Reference: X. Tian, J. Wang, H. Xu, E. S. Chng, and H. Li, “Average modeling approach to voice conversion with non-parallel data,” in Proceedings of the Odyssey 2018. ISCA, June 2018, pp. 227–232.
   author = {Tian, Xiaohai and Wang, Junchao and Xu, Haihua and Chng, Eng Siong and Li, Haizhou},
   title  = {Average Modeling Approach to Voice Conversion with Non-Parallel Data},  
   booktitle = {Proceedings of the Odyssey 2018}, 
   year  = {2018}, 
   month = {June}, 
   pages = {227-232}, 
   location = {Les Sables d'Olonne, France},
   publisher = {ISCA},