Development of Speech Synthesis System for Russian Language Using Deep Neural Networks

Student: Vovk Ivan

Faculty: HSE Tikhonov Moscow Institute of Electronics and Mathematics (MIEM HSE)

Educational Programme: Information Science and Computation Technology (Bachelor)

Year of Graduation: 2019

Current work illustrates a speech synthesis system for Russian language based on neural approaches entirely. Our proposed architecture combines the insights of modified implementations: regularized Tacotron 2 and noise shaping vocoder WaveNet. Thus, it is capable to generate directly from any text given as an input high-fidelity natural speech that can be barely distinguished from human. Specifically, it achieves a mean opinion score of 4,37. Furthermore, we payed a lot of attention to dataset collecting and its preprocessing with the help of personally developed stress detecting algorithms using approximated word embeddings learned on char-level. To our best knowledge, they deliver the accentuation accuracy as good as the best publicly available solutions based on deep neural networks.

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses