The doctoral dissertation in the field of Computer Science will be examined at the Faculty of Science, Forestry and Technology, Joensuu campus and online.
What is the topic of your doctoral research? Why is it important to study the topic?
My doctoral research is computational modelling of emotional responses in music. It focuses on how listeners perceive emotions in music, how musical features relate to emotional responses and how artificial intelligence (AI) can recognize and generate emotionally expressive music.
This topic is important because music plays a significant role in people’s daily lives, including emotion regulation, therapy, and entertainment. However, most research on music and emotions often relies on Western-centric datasets, variations in the modelling of emotion, as well as the simplified representations of emotional experiences. Additionally, there is a limited understanding of how low-level acoustic features influence emotional responses to music.
My research bridges this gap by integrating knowledge from music psychology and computer science to develop data-driven methods that enhance the understanding, recognition, and generation of emotions in music.
What are the key findings or observations of your doctoral research?
This thesis demonstrates that emotional responses to music can be effectively modelled using computational and machine learning. One key finding is that music often evokes mixed emotions in listeners, showing that emotional responses to music are complex and subjective. To address conceptual inconsistencies in the field, this study proposes a unified definition of affective algorithmic composition (AAC) systems.
The study also introduces Emotify+, a culturally diverse multi-label music emotion dataset. Emotify+ dataset provides cultural diversity and broader emotional coverage, making it a valuable resource for research on music emotion recognition. It also shows that low-level acoustic features such as spectral flux, spectral roll-off, and spectral centroid are significantly associated with perceived emotions. Furthermore, clustering analysis shows that emotional responses are naturally grouped into similar emotional patterns rather than being completely distinct from one another.
For the scientific community, this study presents a culturally diverse benchmark dataset (Emotify+), a unified AAC definition, and empirical evidence connecting musical features to emotional perceptions.
For the general public, these findings improve the understanding of how music influences emotions and how AI can interpret these responses. This work can be applied in music recommendation, therapy, and adaptive music systems that respond to users’ emotional states in real-time.
How can the results of your doctoral research be utilised in practice?
The methods and results offer valuable insights for researchers, music technologists, and industry practitioners in analyzing and understanding how musical features influence different emotions.
In practice, these findings can support the development of music recommendation systems that suggest songs based on users’ emotional preferences and reactions. Additionally, the multi-label emotional model can be utilized in interactive and adaptive video games to dynamically generate or adjust background music in real-time according to a player’s emotions, actions, and choices.
The findings can also support creative industries by enabling affective algorithmic composition (AAC) systems to generate musical pieces that match the intended emotional goals of films, games, and multimedia content. In summary, this study advances our understanding of how music influences emotions and provides a foundation for intelligent music technologies that can adapt to various emotional experiences.
What are the key research methods and materials used in your doctoral research?
The study employed a computational and experimental approach that involved systematic review, dataset development, machine learning, statistical analysis, and clustering techniques. First, a systematic review was conducted on affective systematic composition (AAC) studies following established review guidelines to identify existing methods, emotional models, machine learning techniques, and research gaps in the field.
Subsequently, Emotify +, a multi-label music emotion dataset containing 400 musical excerpts from four different genres was created. Low-level acoustic features related to rhythm, timbre, dynamics, and spectral properties were extracted using the jMIR tool. Supervised learning methods and correlation analysis were performed to examine the relationships between acoustic features and perceived emotions. Finally, clustering analysis was performed to identify emotional groupings and similarities in the music emotion profiles.
Is there something else about your doctoral dissertation you would like to share in the press release?
I would like to highlight the importance of the Emotify + dataset. It discusses cultural diversity in music emotion research, as emotional responses to music are often studied using datasets that are predominantly Western-centric. By incorporating participants from different cultural backgrounds, this study provides a broader insight into how people perceive emotions in music.
This study also highlights the significance of moving beyond rigid emotional categories to multi-label emotional modelling, which more accurately reflects real human emotional experiences with music. Finally, the findings may contribute to the development of emotion-aware technologies that use music to enhance well-being, entertainment, and personalize digital experiences.
The doctoral dissertation of Abigail Wiafe, MSc, entitled Computational modelling of emotional responses in music will be examined at the Faculty of Science, Forestry and Technology, Joensuu campus and online. The opponent will be University Lecturer Antti Laaksonen, University of Helsinki, and the custos will be Professor Pasi Fränti, University of Eastern Finland. Language of the public defence is English.
For further information, please contact:
Abigail Wiafe, [email protected]
