Music.Me: A Personalised Music Recommendation System for Wellbeing
Project Overview
This project enhances Music.Me, a personalised music recommendation platform, by incorporating emotion recognition into the recommendation process. While most systems rely only on listening history, we explored how music influences emotions and well-being by predicting two key affective dimensions: arousal (energy) and valence (positivity). Our approach was inspired by the 2025 paper “Towards Unified Music Emotion Recognition across Dimensional and Categorical Models”. We replicated and fine-tuned their multitask learning framework, adding additional datasets to improve robustness. Audio features such as spectrograms, embeddings, and chord progressions were used, with experiments also planned for OpenSMILE and Essentia.
What were the key results of your research project?
Results show that the unified model achieves strong predictive performance, outperforming baseline methods. This paves the way for recommendations tailored not only to user preferences but also to their emotional state and daily activities, supporting music’s role in everyday well-being.
Download presentation slides
How do you feel you have benefited from completing this internship, and has it made you consider future career paths?
Taking part in the N8CIR Internship has given me hands-on experience in applying machine learning to a real-world challenge, strengthening both my technical and research skills. It also helped me develop confidence in communicating complex ideas to a broad audience. This experience has confirmed my interest in pursuing a career in AI research, particularly in projects that connect technology with human well-being.
Download presentation slides
Internships 2025 - Al Aiham Ahmed