@inproceedings{oai:nitech.repo.nii.ac.jp:00003450,
 author = {Wu, Yi Jian and King, Simon and Tokuda, Keiichi},
 book = {Proceedings - 2008 6th International Symposium on Chinese Spoken Language Processing, ISCSLP 2008},
 month = {Dec},
 note = {application/pdf, This paper explores a cross-lingual speaker adaptation technique forHMM-based speech synthesis, where a source voice model for Englishis transformed into a target speaker model using MandarinChinese speech data from the target speaker. A phone mappingbasedmethod is adopted to map Chinese Initial/Finals into Englishphonemes and two types of mapping rules, including one-to-one andone-to-sequence mappings, are compared. In order to avoid havingto map prosodic features between languages, the adaptation procedureuses regression classes and transforms that are constructedfor triphone models, then used to adapt the phonetic-and-prosodiccontext-dependent models. From the experimental results, we foundthat a one-to-sequence phone mapping is better than a one-mapping, and that the similarity between adapted English and target Chinese speaker is reasonable.},
 pages = {9--12},
 publisher = {Institute of Electrical and Electronics Engineers},
 title = {Cross-Lingual Speaker Adaptation for HMM-Based Speech Synthesis},
 year = {2008}
}