//Personalised music selections effecting emotion and mood

Personalised music selections effecting emotion and mood


Research study which involved development of client and server applications to curate personalised music playlists based on predictive computational models.

Jump to Chapter 4

I was involved in multiple aspects of this study:
(1) Personalised music selections for driving scenarios
(2) Personalised music selections for mitigating the symptoms of depression
(3) Personalised music selection online service

Personalised music selections for driving scenarios

Development of applications to facilitate driving experiments to investigate the impact of listening to music while driving and driving behaviour.

The ‘flow drive’ application was used to collect data on driving behaviour (see below).

The software developed as part of this project is contained with the follow two version controlled repositories.

Polyhymnia Driving Application Gitrepo: music-driving

Polyhymnia Android Application Gitrepo: polyhymnia-player-android

Personalised music selections for mitigating the symptoms of depression

Since low mood is a critical symptom of depressive disorders, music emerges as a potential therapeutic tool due to its effectiveness for the induction of emotional and mood states. Music is commonly used by people in everyday life to regulate affective states, and it has demonstrated effects in mood regulation both in clinical and non-clinical populations

Music player application was developed which could help to influence mood based on computational model to control personalised playlists.

A simple example of how a customised playlists is generated is explained here.
The starting point, which is input into the system is the users current mood state, and the target mood state. In this example, two characteristics have been used, (1) arousal and (2) valence.

The plot above shows the users current state (red triangle bottom left) and the target mood state (green square top right). The blue line indicates the trajectory from the current users state to the target state.

The plot below shows the same information as the plot above with the addition of segment points, which are shown as blue boxes on the trajectory line (seg 0, seg 1 and seg 2).

These segment points represent the number of music tracks that will be curated in a users playlist. In this case there are 3 music tracks that will be used to ‘move’ the user from these current state to the target state. As shown, these segmentation points are equally spaced between the user and target states. These segmentation points are used to discover the ‘closest’ music track to that point.

The plot above shows the user and target states with the same blue line indicating the trajectory between the two states. This plot also shows the music tracks in terms of arousal and valence values. In this example there are 4 music tracks.

The final plot above shows all elements described above, with the additional information of the music tracks which has been selected as ‘closest’ to the segmentation points. This list of tracks is effectively the ordered playlist which has a trajectory from the users current mood state to the target mood state.

Two separate applications were developed to support this work.

The first was a desktop application which supported local music library of tracks. The software repository for this application: Polyhymnia Server Desktop Application Gitrepo: polyhymnia-local.

The second application developed was integrated with the online music service Spotify such that a users music selection could be derived from favourite tracks or genre of music. The software repository for this application: Polyhymnia Spotify Application Gitrepo: polyhymnia-spotify.

Personalised music selection online service

This section provides information on the online service which was developed as part of a clinical study.

The software repository for this application: Polyhymnia Spotify Web Server Application Gitrepo: polyhymnia-server.

Software

This section provides links to the version controlled repositories that were used for the array of software developed during this study.

Polyhymnia Driving Application
Gitrepo: music-driving

Polyhymnia Android Application
Gitrepo: polyhymnia-player-android

Polyhymnia Server Desktop Application
Gitrepo: polyhymnia-local

Polyhymnia Spotify Application
Gitrepo: polyhymnia-spotify

Polyhymnia Spotify Web Server Application
Gitrepo: polyhymnia-server.

Further Reading

Below is a list of related code snippets and other resources used as part of the study that might be of interest.

  • http://ayeshalshukri.co.uk/technical-guides/how-to-install-tensor-flow-on-ubuntu/
  • http://ayeshalshukri.co.uk/technical-guides/how-to-install-tflearn/