Music Recommender Systems Based on Physiological Signals
![](/project/music-emotion-recognition/featured_hu7fba9da1679ce4a56c592454604cb9c1_461667_720x0_resize_q90_lanczos.jpg)
The state-of-the-art emotion-based music information retrieval (MIR) systems have focused on modelling music emotion from acoustic music features. This project aims to further personalize emotion-aware music recommendation via incorporating user-related modalities (i.e., physiological signals).
In this project, we conducted a user experiment that simulated a real-life music discovery scenario. The dataset built from this experiment will be used to build the music emotion recognition (MER) module of the music recommendation system.
Highlights:
- Designed and performed a user experiment to build a dataset with synchronized physiological signals (BVP, HR, IBI, EDA, TEMP) and user-labelled music-induced emotion;
- Built the music emotion recognition (MER) model using physiological features and music features.