A Conversion Method of Lhasa Tibetan Tones Based on Pitch Target Model

摘要

Using the method of conversion prosodic features we can realize high quality speech synthesis. Lhasa Tibetan, a central Tibetan language, is a register tonal language .Firstly, a Pitch Target model which is optimized to describe Lhasa Tibetan FO contours is introduced; based on this, a conversion method of Lhasa Tibetan is proposed. We can convert one tone FO contour to another by this method .Using the converted FO contour,we can generate high quality Lhasa Tibetan speech by STRAIGHT algorithm. On the other hand, the required capacity of speech database for speech synthesis can be reduced. Subjective experiments demonstrated that the converted monosyllables achieve 4.0 of the average mean opinion score(MOS).