Thesis Details

Vliv akustiky prostředí na úspěšnost rozpoznávače řeči

Master's Thesis Student: Paliesek Jakub Academic Year: 2020/2021 Supervisor: Szőke Igor, Ing., Ph.D.
English title
Impact of Environment Acoustics on Speech Recognition Accuracy
Language
Czech
Abstract

This diploma thesis deals with impact of room acoustics on automatic speech recognition (ASR) accuracy. Experiments were evaluated on speech corpus LibriSpeech and database of impulse responses and noise called ReverbDB. Used ASRs were based on Mini LibriSpeech recipe for Kaldi. First it was examined how well can ASR learn to transcribe in selected environments by using the same acoustic conditions during training and testing. Next, experiments were carried out with modifications of ASR architecture in order to achieve better robustness against new conditions by using methods for adapation to room acoustics - r-vectors and i-vectors. It was shown that recently proposed method of r-vectors is beneficial even when using real impulse responses for data augmentation.

Keywords

speech recognition, room acoustics, adaptation, impulse response

Department
Department of Computer Graphics and Multimedia FIT BUT
Degree Programme
Information Technology, Field of Study Information Technology Security
Files
Status
defended, grade C
Date
23 June 2021
Reviewer
Karafiát Martin, Ing., Ph.D.
Committee
Drahanský Martin, prof. Ing., Dipl.-Ing., Ph.D. (DITS FIT BUT), předseda
Hrubý Martin, Ing., Ph.D. (DITS FIT BUT), člen
Malinka Kamil, Mgr., Ph.D. (DITS FIT BUT), člen
Očenášek Pavel, Mgr. Ing., Ph.D. (DIFS FIT BUT), člen
Vojnar Tomáš, prof. Ing., Ph.D. (DITS FIT BUT), člen
Citation
PALIESEK, Jakub. Vliv akustiky prostředí na úspěšnost rozpoznávače řeči. Brno, 2021. Master's Thesis. Brno University of Technology, Faculty of Information Technology. 2021-06-23. Supervised by Szőke Igor. Available from: https://www.fit.vut.cz/study/thesis/23934/
BibTeX 
@mastersthesis{FITMT23934,
    author = "Jakub Paliesek",
    type = "Master's thesis",
    title = "Vliv akustiky prost\v{r}ed\'{i} na \'{u}sp\v{e}\v{s}nost rozpozn\'{a}va\v{c}e \v{r}e\v{c}i",
    school = "Brno University of Technology, Faculty of Information Technology",
    year = 2021,
    location = "Brno, CZ",
    language = "czech",
    url = "https://www.fit.vut.cz/study/thesis/23934/"
}

Theses

Back to top