Yong Zhao

TL;DR: In this article, a text processing system for processing multi-lingual text for a speech synthesizer includes a first language dependent module for performing at least one of text and prosody analysis on a portion of input text comprising first language.

TL;DR: In this article, a method and apparatus are provided for refining segmental boundaries in speech waveforms, where contextual acoustic feature similarities are used as a basis for clustering adjacent phoneme speech units.

TL;DR: In this article, a concatenating speech synthesizer concatenates selected speech units to obtain the desired synthesized speech by selecting replacement speech units based on measures representative of the difference between the HMM acoustic models of the desired speech unit and available speech units.

TL;DR: In this paper, a prosody model is used to determine whether the synthesized speech will sound unnatural, and if so, the speech is regenerated with modified data and the evaluation and regeneration may be iterative until deemed natural sounding.

TL;DR: A new type of End-to-End system for text-dependent speaker verification is presented, using a speaker discriminate CNN to extract the noise-robust frame-level features and the proposed attention model takes the speaker discriminate information and the phonetic information to learn the weights.