Jun 25, 2019   11:51 a.m. Tadeáš
Academic information system

Summary of topics offered - Slovak university of technology in Bratislava


Basic information

Type of work: Dissertation thesis
Topic: Hlboké neurónové siete v end-to-end rozpoznávaní reči
Title of topic in English: Deep neural networks in end-to-end speech recognition
State of topic: approved (prof. RNDr. Otokar Grošek, PhD. - Chairperson of Departmental Board)
Thesis supervisor: Ing. Milan Rusko, PhD.
Faculty: Faculty of Informatics and Information Technologies
Supervising department: Faculty of Informatics and Information Technologies
Max. no. of students: --
Academic year:2019/2020
Proposed by: Ing. Milan Rusko, PhD.
External educational institution: Institute of Informatics of the Slovak Academy of Sciences
Annotation: Uvedenie Hlbokých neurónových sietí do systémov rozpoznávania reči podstatne zlepšilo spoľahlivosť rozpoznávačov a ich odolnosť voči zmenám v charakteristikách hlasu hovoriaceho a prenosového kanála. Doktorand vo svojej práci podá prehľad o tomto novom prístupe a jeho použití vo svete vrátane end-to-end prístupov. S využitím existujúcich rečových a textových databáz vyvinutých na UI SAV navrhne a zrealizuje end-to-end rozpoznávač reči v slovenčine a porovná ho s predošlými prístupmi.
Annotation in English: The introduction of Deep Neural Networks into speech recognition systems has significantly improved the reliability of the recognizers and their resistance to inter-speaker and intra-speaker variability, and various transmission channel characteristics. The PhD student will give an overview of this new approach and its use in the speech recognition, including end-to-end approaches. Using existing speech and text databases developed at UI SAV, he will design and implement the end-to-end speech recognizer in Slovak and compare it to the previous approaches.



Limitations of the topic

To sign up for a topic it is necessary to fulfil one of the following restrictions

Limit to study programme
The table shows limitations of study programme, field, track the student has to be enrolled in to be able to register for a given topic.

ProgrammeTrackTrack
D-AI Applied Informatics -- not entered -- -- not entered --
D-AIxA Applied Informatics (study in English)-- not entered -- -- not entered --
D-AI4 Applied Informatics -- not entered -- -- not entered --
D-AI4xA Applied Informatics (study in English)-- not entered -- -- not entered --