CONF Liang_INTERSPEECH_2011/IDIAP Phonological Knowledge Guided HMM State Mapping for Cross-Lingual Speaker Adaptation Liang, Hui Dines, John cross-lingual speaker adaptation HMM-based TTS minimum generation error phonological knowledge EXTERNAL https://publications.idiap.ch/attachments/papers/2011/Liang_INTERSPEECH_2011.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/Liang_Idiap-RR-17-2011 Related documents Proceedings of Interspeech Florence, Italy 2011 Within the HMM state mapping-based cross-lingual speaker adaptation framework, the minimum Kullback-Leibler divergence criterion has been typically employed to measure the similarity of two average voice state distributions from two respective languages for state mapping construction. Considering that this simple criterion doesn't take any language-specific information into account, we propose a data-driven, phonological knowledge guided approach to strengthen the mapping construction -- state distributions from the two languages are clustered according to broad phonetic categories using decision trees and mapping rules are constructed only within each of the clusters. Objective evaluation of our proposed approach demonstrates reduction of mel-cepstral distortion and that mapping rules derived from a single training speaker generalize to other speakers, with subtle improvement being detected during subjective listening tests. REPORT Liang_Idiap-RR-17-2011/IDIAP Phonological Knowledge Guided HMM State Mapping for Cross-Lingual Speaker Adaptation Liang, Hui Dines, John EXTERNAL https://publications.idiap.ch/attachments/reports/2011/Liang_Idiap-RR-17-2011.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/Liang_INTERSPEECH_2011 Related documents Idiap-RR-17-2011 2011 Idiap June 2011 Within the HMM state mapping-based cross-lingual speaker adaptation framework, the minimum Kullback-Leibler divergence criterion has been typically employed to measure the similarity of two average voice state distributions from two respective languages for state mapping construction. Considering that this simple criterion doesn't take any language-specific information into account, we propose a data-driven, phonological knowledge guided approach to strengthen the mapping construction -- state distributions from the two languages are clustered according to broad phonetic categories using decision trees and mapping rules are constructed only within each of the clusters. Objective evaluation of our proposed approach demonstrates reduction of mel-cepstral distortion and that mapping rules derived from a single training speaker generalize to other speakers, with subtle improvement being detected during subjective listening tests.