Emotion Recognition in Javanese Music: A Comparative Study of Classifier Models with a Human-Annotated Dataset

Septianto, Moh Erwin
Purnomo, Ariana Tulus
Lin, Ding Bing
Kim, Chang Soo

Indonesian Journal of Computing, Engineering, and Design
Vol. 7 No. 2 (2025): IJoCED

With advancements in machine learning and the increasing availability of music datasets, Music Emotion Recognition (MER) has gained significant attention. However, research focusing on Indonesian traditional music, particularly Javanese music, remains limited. Understanding emotions in Javanese music is crucial for preserving cultural heritage and enabling emotion-aware applications tailored to Indonesian traditional music. This study investigates the effectiveness of three well-established machine learning models, 1D Convolutional Neural Networks (1D-CNNs), support Vector Machines (SVMs), and XGBoost, in classifying emotions in Javanese music using a manually annotated dataset. The dataset consists of 100 Javanese songs from various genres, including Dangdut, Koplo, and Campur Sari, annotated based on the Thayer emotion model. The models’ performance was assessed using different data split ratios, with accuracy rates exceeding 70%. Among the tested classifiers, SVM exhibited the highest and most stable accuracy.