Rekonstrukcija zvučnih zapisa primjenom metoda strojnog učenja

master's thesis

• Title (english): Reconstruction of Audio Recordings Using Machine Learning Techniques
• Author: Lovro Kordiš
• Mentor: Goran Delač (mentor)
• Committee member: Goran Delač (predsjednik povjerenstva)
• Committee member: Marin Šilić (član povjerenstva)
• Committee member: Siniša Srbljić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-09, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Rekonstrukcija zvuka područje je dubokog učenja koje je još uvijek u razvoju. Ovaj se rad bavi generativnim suparničkim mrežama i korištenjem suparničkog treninga na parovima zvuka niske i visoke razlučivosti kako bi se stvorio model za rekonstrukciju zvuka. Detaljno su opisane duboke konvolucijske i generativnesuparničke mreže. Implementiran je duboki konvolucijski U-Net suparnički model, te su ispitane dvije funkcije gubitka - Wasserstein i Relativistički gubitak. Model je evaluiran na VCTK skupu podataka. Na kraju su prikazani rezultati i slike spektrograma.

Abstract (english)

Audio super resolution is still a developing field in deep learning. This paper deals with generative adversarial networks and using the adversarial training on pairs of low and high resolution audios to create a model for audio super resolution. Deep convolutional and generative adversarial networks are described in detail. A deep convolutional U-Net adversarial model is implemented, and two adversarial loss functions are tested - Wasserstein and Relativistic loss. Model is evaluated on VCTK dataset. Finally, the results and spectogram images are presented.

• Keywords: strojno učenje; duboko učennje; generativne suparničke mreže; rekonstrukcija zvuka
• Keywords (english): machine learning; deep learning; generative adversarial networks; audio super resolution
• Language: english
• URN:NBN: urn:nbn:hr:168:954910
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2020-01-07 19:12:57