Change search
ReferencesLink to record
Permanent link

Direct link
Developing a soprano classifier using FIR-ELM neural network
Mälardalen University, School of Innovation, Design and Engineering.
2016 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

This thesis aims at investigate the feasibility of classifying the soprano singing voice type using a single layer neural network trained with the FIR-ELM algorithm after that the monaural auditory mixture has been segmented with the Harmonic, Percussive and Residual, HPR, decomposition algorithm, previously introduced by Driedger et al.Two different decomposition structures has been evaluated both based on the same HPR decomposition technique. Firstly one single layer that only take advantage of the result of the more pure harmonic and the more pure percussive components of the signal. Secondly, one multilayer structure that further decompose both the harmonic and the percussive components but also takes into account the components that can not be clearly categorized as neither harmonic or percussive components, these are the residual components. The result of the classification was up to 98.5 $\%$ after using these segmentation techniques, this shows that it is feasibly to classify the singing voice type soprano in an monaural source recorded in a non-professional environment using the FIR-ELM algorithm.

Place, publisher, year, edition, pages
2016. , 29 p.
Keyword [en]
FIR-ELM, Audio decomposition, Soprano classifier
National Category
Engineering and Technology
URN: urn:nbn:se:mdh:diva-33399OAI: diva2:1034666
Subject / course
Computer Science
Available from: 2016-10-13 Created: 2016-10-12 Last updated: 2016-10-13Bibliographically approved

Open Access in DiVA

fulltext(2326 kB)10 downloads
File information
File name FULLTEXT01.pdfFile size 2326 kBChecksum SHA-512
Type fulltextMimetype application/pdf

Search in DiVA

By author/editor
Cederblad, Peter
By organisation
School of Innovation, Design and Engineering
Engineering and Technology

Search outside of DiVA

GoogleGoogle Scholar
Total: 10 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

ReferencesLink to record
Permanent link

Direct link