Utomo, Tegar Satrio (2019) Pemetaan Emosi pada Lagu Berdasarkan Russell's Two-Dimensinal Model Menggunakan Gaussian Process. Other thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
05111540000178-Undergraduate_Theses.pdf Download (3MB) | Preview |
Abstract
Dalam Music information retriaval, terdapat beberapa hal yang lazim dilalukan seperti artist identification, instrument recognition dan emotion recognition. Salah satu pengaplikasian Music information retriaval adalah emotion recognition. Emotion recognition pada musik dapat dilakukan salah satunya dengan memanfaatkan model Russell, yaitu sebuah bidang dalam sumbu x (valence) dan y (arousal) yang membagi emosi ke dalam model dimensional representation yang memetakan emosi kedalam cluster - cluster yang merepresentasikan kelompok emosi yang memiliki sifat sama. Pada tugas akhir ini akan dicari nilai valence - arousal untuk model Russell yang merepresentasikan emosi dari suatu lagu dengan gaussian process regression. Data pelatihan dan uji coba diambil dari dataset MediaEval’2013 yang merupakan data yang berisi lagu - lagu bebas hak cipta dan memiliki anotasi valence (sumbu x) dan arousal (sumbu y) yang nantinya menjadi tujuan pemetaan lagu pada bidang emosi. Akan dibuat dua model yang digunakan untuk memprediksi nilai untuk valence (sumbu x) dan arousal (sumbu y). Setiap model akan melalui tahap Preprocessing, antara lain penyeragaman sampling rate, serta perubahan format lagu menjadi wav, yang kemudian akan diambil fitur fitur seperti mel frequency cepstral coefficients, timbre features, spectral crest factor dan spectral flatness measure, chromagram. Akan digunakan metode Gaussian process regressor dengan fungsi kernel Rational Quadratic Hasil uji coba terakhir didapatkan nilai R2 0.68 untuk model arousal dan 0.38 untuk model valence, serta akurasi pemetaan sebesar 63%.
================================================================================================================================
In Music information retriaval, there are some things that are commonly practiced such as artist identification, instrument recognition and emotion recognition. One application of Music information retriaval is emotion recognition. Emotion recognition in music can be done by using the Russell model, which is a field in the valence (x axis) and arousal (y axis) that divides emotions into dimensional representation models that map emotions into clusters - clusters that represent groups of emotions that have the same nature . In this final project, the value of valence - arousal for the Russell model that represents the emotion of a song with gaussian process regression will be sought. Training data and trials were taken from the MediaEval '2013 dataset which is data that contains copyright-free song songs and has an annotation of the x-axis and y-axis which later becomes the goal of song mapping in the emotional field. Two models will be used to predict values for the valence (x axis) and arousal (y axis). Each model will go through the Preprocessing stage, including sampling rate uniformity, and changes to the song format to wav, which will then take features such as mel frequency cepstral coefficients, timbre features, spectral crest factor and spectral flatness measure, chromagram. Gaussian process method will be used regressor with Rational Quadratic kernel function The results of the last trial obtained R2 0.68 for the arousal model and 0.38 for the valence model, and 63% for mapping accuracy.
| Item Type: | Thesis (Other) |
|---|---|
| Additional Information: | RSIf 005.74 Uto p-1 2019 |
| Uncontrolled Keywords: | Dataset MediaEval’2013, Emotion Recognition, Gaussian Process Regression, Music Information Retriaval, Model Russell |
| Subjects: | Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression T Technology > T Technology (General) > T57.5 Data Processing |
| Divisions: | Faculty of Information Technology > Informatics Engineering > 55201-(S1) Undergraduate Thesis |
| Depositing User: | Tegar Satrio Utomo |
| Date Deposited: | 11 Jun 2024 02:39 |
| Last Modified: | 11 Jun 2024 02:39 |
| URI: | http://repository.its.ac.id/id/eprint/65889 |
Actions (login required)
 |
View Item |