diplomsko delo
Tom Šabanov (Author), Marko Robnik Šikonja (Mentor)

Abstract

V diplomskem delu smo naslovili problem sinteze slovenskega govora na podlagi sorazmerno majhne učne množice. Opisali smo starejše pristope sinteze govora, kot sta artikularna in formantna sinteza, ter sodobne pristope sinteze z združevanjem enot in sinteze govora s pomočjo globokih nevronskih mrež. Ustvarili smo različne podatkovne množice iz 30 ur govora štirih govorcev, ki smo jih uporabili za sintezo govora. Uporabili smo arhitekturi ForwardTacotron za generiranje mel-spektrogramov ter Hifi-GAN za pretvorbo teh spektrogramov v zvočne signale. Ustvarili smo splošni model za moški govor, ki ga je možno prilagoditi na nove govorce. Najboljši ustvarjeni sistem dosega dobro povprečno oceno poslušalcev (4.07 na lestvici od 1-5) in daje vtis naravnega govora.

Keywords

sinteza slovenskega govora;globoke nevronske mreže;model Tacotron;računalništvo in informatika;univerzitetni študij;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [T. Šabanov]
UDC: 004.8:81'322(043.2)
COBISS: 75236355 Link will open in a new window
Views: 326
Downloads: 74
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Slovene speech synthesis using multi-speaker datasets
Secondary abstract: In the thesis, we addressed the problem of Slovene speech synthesis based on relatively small data set. We described older approaches to speech synthesis like articular and formant synthesis, and more modern approaches like unit selection and speech synthesis with deep neural networks. We created a dataset consisting 30 hours of speech from four speakers for use with speech synthesis. We used ForwardTacotron architecture for generating mel-spectrograms and Hifi-GAN architecture for generating waveforms from these spectrograms. We created a basic model for male speech, which can be fine-tuned for new speakers. The best system we created achieved a good mean opinion score of listeners (4.07 on a scale 1-5) that simulates natural speech.
Secondary keywords: Slovene speech synthesis;deep neural networks;Tacotron model;computer and information science;diploma;Računalniško jezikoslovje;Umetna inteligenca;Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS): Bachelor thesis/paper
Study programme: 1000468
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 40 str.
ID: 13296241