USING ANIMATED MOOD PICTURES IN MUSIC RECOMMENDATION - Information Visualisation: Techniques, Usability and Evaluation - page 223

object to the dashboard or customize the car properties to look more like a
sports car. This would then have an effect on the seed songs associated
with the picture.
Several users were impressed by the animated elements added to the
static images, and they hoped to see animated actions in future versions of
the prototype. Using animated actions would clearly point out the most
important message of the picture, and it could bring out the intended mood
more clearly.
Conclusions and future work
In this chapter, we have studied the idea of using animated mood pictures
in the context of music recommendation. In the implemented software
prototype, the user interacted with a collection of pictures to receive new
music recommendations from the associated genres.
In the initial interview, the idea of selecting music based on mood
pictures was found to be very good by 85% of the participants. The
participants also felt that the prototype looked good, was easy and fun to
use, and that it presented music in a novel way.
In longer-term use, the prototype lacked some important features,
which had a clear effect on the longer-term ratings. Most importantly, a
majority of the participants would have wanted to add their own pictures
and music to the system as well as to change the picture and music
associations of the built-in pictures. One reason for this wish was that it
can be extremely hard to find such default mood picture/seed song pairs to
please every user.
Despite the flaws, 60% of the participants said that they could use the
prototype, as it is, to complement their other music player applications.
This indicates that the concept of associating music to mood pictures is a
promising way of accessing music collections and exploring new music.
The concept could easily be scaled down to fit the screen of modern
mobile devices, which include high-quality built-in cameras and dedicated
music player applications, and which people carry around all the time. The
pictures used were designed to match the potential daily situations and
activities of the target user group. While the variation in the grades of
match between pictures and music can be partially explained by
differences in musical taste, it may also indicate that the pictures were too
ambiguous. Thus, the next concrete step is to redefine the mood pictures to
more clearly represent discrete moods, and select a small set of songs that
could be matched to these pictures in a user study. This would help to
clarify whether there is a clear link between the mood pictures designed