Related Concept Videos
Labeling Emotion
Motional Emf
Self-Report Tests of Personality
You might also read
Related Articles
Articles linked to this work by shared authors, journal, and citation graph.
Development and User Experiences of a Novel Virtual Reality Task for Poststroke Visuospatial Neglect: Exploratory Case Study.
Personalized Interactive Music Systems for Physical Activity and Exercise: Exploratory Systematic Review and Meta-Analysis.
Related Experiment Video
Updated: Aug 15, 2025

Examining Changes in HRV and Emotion Following Artmaking with Three Different Art Materials
Published on: January 11, 2020
MERP: A Music Dataset with Emotion Ratings and Raters' Profile Information.
En Yan Koh1, Kin Wai Cheuk1, Kwan Yee Heung1
1Information Systems Technology and Design Pillar, Singapore University of Technology and Design, Singapore 487372, Singapore.
This study introduces the Music Emotion Recognition with Profile information (MERP) dataset, featuring dynamic emotional ratings for 54 songs. The MERP dataset includes music and user profile data to advance music emotion recognition research.
Area of Science:
- Computer Science
- Music Information Retrieval
- Affective Computing
Background:
- Music's emotional impact is subjective and varies greatly among listeners.
- Existing datasets lack comprehensive music features and detailed user profile information for emotion analysis.
- Understanding music-induced emotions requires considering both acoustic properties and listener characteristics.
Purpose of the Study:
- To introduce and describe the novel Music Emotion Recognition with Profile information (MERP) dataset.
- To provide a rich dataset with dynamic valence-arousal ratings, music features, and annotator profile information.
- To establish baseline models for music emotion recognition using the MERP dataset.
Main Methods:
- Collected dynamic valence and arousal ratings for 54 songs via Amazon Mechanical Turk (MTurk).
- Selected songs using a Triple Neural Network and OpenSmile toolkit to cover the valence-arousal space.
- Incorporated annotator demographic, listening preference, and musical background data.
- Developed baseline emotion prediction models using fully connected and LSTM networks.
Main Results:
- The MERP dataset contains detailed annotations for 54 songs, with data from 277 cleaned participant profiles.
- Songs were strategically chosen to span the four quadrants of the valence-arousal space.
- Baseline models demonstrated the feasibility of predicting music-induced emotions using the MERP dataset.
Conclusions:
- The MERP dataset offers a valuable resource for advancing research in music emotion recognition.
- Integrating user profile information alongside music features enhances the understanding of subjective emotional responses.
- The established baseline models provide a foundation for future, more sophisticated music emotion recognition systems.

