@inproceedings{oai:nitech.repo.nii.ac.jp:00003399,
 author = {Wu, Yi-Jian and Tokuda, Keiichi},
 book = {INTERSPEECH 2008 9th Annual Conference of the International Speech Communication Association},
 month = {},
 note = {application/pdf, A minimum generation error (MGE) criterion had been proposedto solve the issues related to maximum likelihood (ML)based HMM training in HMM-based speech synthesis. In thispaper, we improve the MGE criterion by imposing a log spectraldistortion (LSD) instead of the Euclidean distance to define thegeneration error between the original and generated line spectralpair (LSP) coefficients. Moreover, we investigate the effect ofdifferent sampling strategies to calculate the integration of theLSD function. From the experimental results, using the LSDscalculated by sampling at LSPs achieved the best performance,and the quality of synthesized speech after the MGE-LSD trainingwas improved over the original MGE training., Brisbane, AustraliaSeptember 22-26, 2008},
 pages = {577--580},
 publisher = {International Speech Communication Association},
 title = {Minimum generation error training with direct log spectral distortion on LSPs for HMM-based speech synthesis},
 year = {2008}
}