Abstract : | The purpose of this study is to examine through a listening experiment and relevant theoretical approaches, whether emotions, evoked by music stimuli, can be considered as implicit rating data. Five subjects participated on the experiment in which, ten music excerpts of different valence were played to each individual separately. They had to rate the music they listened to, so as to provide us with their preferences towards each excerpt. Furthermore, the subjects while listening to music, had also their physiological measurements of galvanic skin response (GSR), skin temperature (ST) and beats per minute (BPM) recorded, through specific sensors implemented on a wristband. These particular measures were chosen because of the fact that, as stated in other studies, they have a strong relationship with the valence and arousal of an emotion and consequently can be used to define it. According to this research, GSR and ST, indeed present a strong correlation with the valence and arousal of subjects’ emotions, as well as with their recorded preferences. Precisely, subjects’ GSR and ST values were quite high during the listening of music they did not like, as opposed to music they liked. Additionally, the calculated correlation between the mentioned measures and ratings, for each individual separately, was negative and strong. Finally, various machine learning algorithms were applied to further investigate the mentioned correlation and make predictions about subjects’ ratings. Although our experiment lacks the appropriate quantity, as the number of participants is low, it is sufficient to provide us with qualitative indications regarding the relationship between individuals’ recorded measures and preferences in form of ratings.
|
---|