Indonesian Music Emotion Recognition Using Rhythm and Timbre Features

Abstract

This research aims to recognize and classify emotion in Indonesian music. Two types
of features were extracted: rhythm and timbre. These features are used to train a
classifier based on Support vector machines. The dataset used in this study consists of 115
Indonesian popular songs and divided into four emotion class: happy, sad, angry, and
relax, which derived from the Thayer’s emotion model. A series of experiments were
conducted with ten repetitions of 10-fold cross-validation and measured by accuracy. The
results show that timbre and rhythm are better for arousal classification than valence
classification.