prikaz prve stranice dokumenta Preslikavanje pjesama u ples temeljeno na neuronskim mrežama
Access restricted to students and staff of home institution
master's thesis
Preslikavanje pjesama u ples temeljeno na neuronskim mrežama
Rijeka: University of Rijeka, Faculty of Engineering, 2021. urn:nbn:hr:190:087160
Baćac, Ivana
University of Rijeka
Faculty of Engineering
Department of Computer Engineering
Section of Software Engineering

Institutional repository: Repository of the University of Rijeka, Faculty of Engineering

Cite this document

Baćac, I. (2021). Preslikavanje pjesama u ples temeljeno na neuronskim mrežama (Master's thesis). Rijeka: University of Rijeka, Faculty of Engineering. Retrieved from https://urn.nsk.hr/urn:nbn:hr:190:087160

Baćac, Ivana. "Preslikavanje pjesama u ples temeljeno na neuronskim mrežama." Master's thesis, University of Rijeka, Faculty of Engineering, 2021. https://urn.nsk.hr/urn:nbn:hr:190:087160

Baćac, Ivana. "Preslikavanje pjesama u ples temeljeno na neuronskim mrežama." Master's thesis, University of Rijeka, Faculty of Engineering, 2021. https://urn.nsk.hr/urn:nbn:hr:190:087160

Baćac, I. (2021). 'Preslikavanje pjesama u ples temeljeno na neuronskim mrežama', Master's thesis, University of Rijeka, Faculty of Engineering, accessed 23 June 2024, https://urn.nsk.hr/urn:nbn:hr:190:087160

Baćac I. Preslikavanje pjesama u ples temeljeno na neuronskim mrežama [Master's thesis]. Rijeka: University of Rijeka, Faculty of Engineering; 2021 [cited 2024 June 23] Available at: https://urn.nsk.hr/urn:nbn:hr:190:087160

I. Baćac, "Preslikavanje pjesama u ples temeljeno na neuronskim mrežama", Master's thesis, University of Rijeka, Faculty of Engineering, Rijeka, 2021. Available at: https://urn.nsk.hr/urn:nbn:hr:190:087160

Metadata
TitlePreslikavanje pjesama u ples temeljeno na neuronskim mrežama
Title (english)Music to dance mapping based on neural networks
AuthorIvana Baćac
MentorGoran Mauša (mentor)
Committee memberSandi Ljubić (predsjednik povjerenstva)
Committee memberIvan Štajduhar (član povjerenstva)
Committee memberGoran Mauša (član povjerenstva)
GranterUniversity of Rijeka
Faculty of Engineering
(Department of Computer Engineering)
(Section of Software Engineering)
Rijeka
Defense date and country2021-09-20, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
U ovom radu demonstrirana je primjena umjetnih neuronskih mreža u rješavanju izazova preslikavanja pjesama na ples. Prikazani su postupci rješavanja navedenog izazova konvolucijskom neuronskom mrežom i višeslojnom mrežom bez povratnih veza. Model konvolucijske neuronske mreže treniran je spektrogramima, a model višeslojne neuronske mreže bez povratnih veza treniran je numericki izraženim glazbenim značajkama: tempom, brojem početnih trenutaka glazbene note ili drugog zvuka i kepstralnim značajkama. Predstavljen je model višekanalske neuronske mreže koji objedinjuje rezultatne modele konvolucijske neuronske mreže i višeslojne neuronske mreže bez povratnih veza. Validacijom navedenih modela zakljuceno je da, s točnosti 84%, najbolje rezultate ostvaruje model višekanalske neuronske mreže. Model konvolucijske neuronske mreže ostvaruje manju, ali još uvijek dobru točnost 73%, a model višeslojne neuronske mreže bez povratnih veza s točnosti 65% ostvaruje slabije rezultate.
Abstract (english)
In this thesis the application of artificial neural networks in solving the challenge of mapping songs to dance is demonstrated. Procedures for solving this challenge with a convolutional neural network and a multilayer network without feedback are presented. The convolutional neural network model was trained by spectrograms, and the multilayer perceptron was trained by numerically expressed musical features: tempo, number of onsets, and cepstrum features. A multi-channel neural network model is presented, which combines trained models of the resulting convolutional neural network and multilayer perceptron. By validating these models it was concluded that, with an accuracy of 84 %, the best results are achieved by the multichannel neural network model. The convolutional neural network model achieves a lower but still good accuracy of 73 %, and the multilayer perceptron model with an accuracy of 65 % achieves poorer results.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:190:087160
Study programmeTitle: Computer Engineering
Study programme type: university
Study level: graduate
Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
Type of resourceText
File originBorn digital
Access conditionsAccess restricted to students and staff of home institution
Terms of useLicence In copyright icon
Created on2021-10-18 09:39:58