Towards Empathic Music Provision for Computer Users

This study explores the automatic provision of music based on a person´s music preferences and activities on a computer. This research presents two classification models: music-general activity and music-specific activity. General activities were classified as leisurely and academic while specific activities as the name of the computer program/application itself. During data-gathering sessions, a test subject was asked to listen to a variety of songs while engaging in different computer activities in a naturalistic manner. An activity-music tracker program logged all the songs played by the user along with the computer application that he was using at that moment as well as the time stamp. The models classified the activity of the user based on the audio features of the songs, which were extracted using JAudio and Music Miner. Classification algorithms were applied using Weka. The algorithm that obtained the highest accuracy was J48. The music-general activity model obtained an accuracy of 84.376% and kappa value of 0.6857 while the music-specific activity model obtained an accuracy of 71.7417% and kappa value of 0.6888. This research serves as part of a separate study of an an empathic music player system that plays songs based on the computer activity, emotion, in the form of electroencephalograph signals (EEG), and past music preferences of a computer user.