| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

Show document Help

Title:Razpoznava slovenskega govora v aplikaciji za osebe z okvaro sluha : magistrsko delo
Authors:ID Kovačič, Blaž (Author)
ID Bošković, Borko (Mentor) More about this mentor... New window
ID Brest, Janez (Comentor)
Files:.pdf MAG_Kovacic_Blaz_2025.pdf (4,29 MB)
MD5: DAECE7AE2620DD4629635B24944F1050
 
Language:Slovenian
Work type:Master's thesis/paper
Typology:2.09 - Master's Thesis
Organization:FERI - Faculty of Electrical Engineering and Computer Science
Abstract:V okviru magistrskega dela smo razvili aplikacijo za osebe z okvaro sluha, ki z razpoznavanjem slovenskega govora omogoča realnočasovno podnaslavljanje predvajanega govora. Za razpoznavanje govora aplikacija uporablja modele Whisper podjetja OpenAI, ki smo jih doučili na superračunalniku VEGA s pomočjo korpusa Artur 1.0. Pri tem smo za primerjavo rezultatov učili dva modela različnih velikosti. Za večji model smo na testni množici dosegli stopnjo napačno razpoznanih besed 11,38 %, medtem ko smo za hitrejši, manjši model dosegli 15,19 %. Realnočasovno izvajanje smo zagotovili z različnimi optimizacijami dekodiranja žetonov in s pomočjo ustreznih zaledij za sklepanje z modeli.
Keywords:razpoznava slovenskega govora, okvara sluha, samodejno podnaslavljanje, strojno učenje
Place of publishing:Maribor
Place of performance:Maribor
Publisher:[B. Kovačič]
Year of publishing:2025
Number of pages:1 spletni vir (1 datoteka PDF (XI, 71 str.))
PID:20.500.12556/DKUM-93585 New window
UDC:004.934(043.2)
COBISS.SI-ID:245832707 New window
Publication date in DKUM:13.08.2025
Views:200
Downloads:39
Metadata:XML DC-XML DC-RDF
Categories:KTFMB - FERI
:
Copy citation
  
Average score:(0 votes)
Your score:Voting is allowed only for logged in users.
Share:Bookmark and Share



Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:CC BY 4.0, Creative Commons Attribution 4.0 International
Link:http://creativecommons.org/licenses/by/4.0/
Description:This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.
Licensing start date:03.07.2025

Secondary language

Language:English
Title:Slovenian speech recognition in an application for people with hearing impairment
Abstract:In this master's thesis, we developed an application for people with hearing loss that enables Slovenian speech recognition for real-time captioning of the device's audio output. The solution uses OpenAI Whisper speech recognition models that we fine-tuned on the VEGA supercomputer using the Artur 1.0 corpus. For comparison of results, we trained two models of different sizes. On the test dataset, we achieved a word error rate of 11.38% for the larger model and 15.19% for the faster, smaller model. We achieved real-time captioning using different token decoding optimizations and appropriate model inference backends.
Keywords:Slovenian speech recognition, hearing loss, automatic captioning, machine learning


Comments

Leave comment

You must log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back
Logos of partners University of Maribor University of Ljubljana University of Primorska University of Nova Gorica