Please use this identifier to cite or link to this item: https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/3686
Title: Melody Analysis for Prediction of the Emotions Conveyed by Sinhala Songs
Authors: Lakshitha, M.G.W
Issue Date: 8-Sep-2016
Abstract: With the intention of addressing the real world phenomena that music composers can compose better melodies which agree with the emotions possessed by the lyrics of songs, we propose a music emotion classi cation solely based on the melodies of music in order to assess the capacity of melodies to di erentiate songs among di erent emotions. if such classi cation is successful, we can stand for the fact that a melody based automated system is able to assist the composers to produce proper melodies which align with the message to be delivered through a particular song. Other than that, a melody based emotion classi cation system can be utilized in situations where a proper indexing is needed for large music databases. The proposed approach extracts the melodies from music excerpts which are 90 seconds long and then some statistical descriptors of the melodies are calculated. Using those descriptors, we assessed the capability of the melodies of Sri Lankan Sinhala music for predicting the emotions possessed by di erent melodies. In order to evaluate the ability of the melody for the intended task, we used several machine learning algorithms with and without feature selection approaches and accuracy enhancement approaches. But no approach gave us a satisfactory level of accuracy for the classi cation. After all the experiments we did using melody features only, the attained inference is that, the melody, in isolation, is not capable of di erentiating di erent emotions conveyed by Sinhala songs. We proved our inference by adding some more features which are non melodic into our feature vector. We added some rhythm features and timbre features and obtained better classi cation results than the results of the earlier classi cations. However it is important to note that, melody is a very important and signi cant aspect for determining the emotions of music, even though it is not capable of doing the job in isolation.
URI: http://hdl.handle.net/123456789/3686
Appears in Collections:SCS Individual Project - Final Thesis (2015)

Files in This Item:
File Description SizeFormat 
MGWLakshitha_11000813.pdf
  Restricted Access
941.41 kBAdobe PDFView/Open Request a copy


Items in UCSC Digital Library are protected by copyright, with all rights reserved, unless otherwise indicated.