Institutional Repository of Vilnius University
Lietuvių kalbos sintezė paslėptų Markovo grandinių metodu /
Abstract
dc:descriptionThe goal of this project was to implement HTK and HTS toolkits to synthesize Lithuanian language. These toolkits use hidden Markov model (HMM) technology. This original goal was successfully achieved. The toolkits can be used to train a model that synthesizes understandable Lithuanian language. This model was used to perform several experiments. During them it was determined that synthesized voice quality greatly depends on Markov chains’ configuration. Finally, the voices generated using this model were compared with voices synthesized using unit selection and neural network methods in human group testing. They had to listen to all voice recordings and evaluate text comprehensibility and acceptability. During these tests it was determined that voices generated using HMM method were neither most acceptable nor most comprehensible in relation to other methods, however overall this method can be used to generate understandable Lithuanian language. The results of this project can be used to improve usage of HTK and HTS toolkits in Lithuanian language synthesis to the point that it is similar to neural network method.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Žiemys, Algirdas,
- Contributors dc:contributor
-
- Kasparaitis, Pijus
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD81917730&prefLang=en_US