Events

We listen to music for enjoyment, distraction, joy, or comfort. The emotions we derive from music listening can have a strong effect on our focus, and therefore our attention and productivity. Although many people often listen to music while they work to help focus, the order and type of music that plays from randomized user libraries is not usually optimized to steer users towards productivity, and can often serve as a distraction instead.

In this project, our goal is to design an intelligent playlist that automatically adjusts the music playing based on the measurements of the user’s attention and productivity levels. Although previous affective music studies create playlists of songs with a similar emotional state, our playlist creation algorithm is designed to take users on an emotional journey.

As a first step towards this goal, we identify a dataset of songs and their associated coordinates in the Arousal-Valence circumplex model (a measure of affective positivity and intensity) and demonstrate the ability to generate a dynamic playlist to bridge the emotional space between a given starting affective state and a target affective state. 

We designed a K-Nearest Neighbors based intelligent algorithm to interpolate between our starting and target affective states. The algorithm generates a path between two emotional states, evaluating neighbors generated from a K-Nearest-Neighbors model using a cosine similarity metric to choose songs.

We found that our model was able to create a smooth playlist of songs from the current affective state to the target destination affective state, creating a direct path through the valence-arousal space. This work is a starting point for further research, as we look to implement ways to integrate our system with facial recognition to examine the user’s emotional state.