Otkrivanje raspoloženja glazbene pjesme korištenjem dubokog učenja

master's thesis

• Title (english): Deep Learning Approach to Detecting the Mood of a Music Song
• Author: Dino Avdić
• Mentor: Marina Bagić Babac (mentor)
• Committee member: Marina Bagić Babac (predsjednik povjerenstva)
• Committee member: Marko Horvat (član povjerenstva)
• Committee member: Krešimir Fertalj (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-13, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Glazba je moćan jezik za izražavanje emocija i raspoloženja kao i sredstvo za utjecaj na emocije slušatelja. Porastom ogromnih digitalnih glazbenih knjižnica pokazalo se da su glazbena raspoloženja postala poželjna pristupna točka glazbenim bazama podataka, a automatsko prepoznavanje i klasifikacija raspoloženja posljednjih godina dobivaju sve veću pozornost. U usporedbi s tradicionalnim karakteristikama glazbe poput žanra ili izvođača, prepoznavanje raspoloženja je subjektivnije i teže ga je kvantificirati, što ga čini izazovnijim. U ovom radu treba prikazati klasifikacijske modele koji mogu odrediti raspoloženja pjesme pomoću dubokog učenja. Potrebno je dizajnirati i implementirati detektor raspoloženja korištenjem trenutno dostupnih skupova podataka.

Abstract (english)

Music is a powerful language to express the emotions and moods as well as a mean to influence emotions of a listener. With the increase of huge digital music libraries over the past decade, music moods have been shown that becoming a desirable access point to music databases, and automatic recognition and classification of mood have received increasing attention in recent years. Compared to traditional characteristics of music such as genre or artist, mood recognition is more subjective and difficult to quantify, which make it more challenging. In this thesis, classification models that can determinate which is the mood of a specific track based on deep learning, are shown. An example of mood detector is designed and implemented using current datasets that are available for this purpose.

• Keywords: duboka neuronska mreža; klasifikacija; glazbene značajke; točnost modela; funkcija gubitka; Adam optimizator; Dropout tehnika
• Keywords (english): deep neural network; classification; musical features; model accuracy; loss function; Adam optimizator; Dropout technique
• Language: croatian
• URN:NBN: urn:nbn:hr:168:086224
• 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: 2023-01-13 11:41:48