Team

Prof. Goffredo Hauss, Andreja Andric, Andrea Fantasia (Università degli Studi di Milano), Pierre-Louis Xech (Microsoft Research Cambridge)

Aim

Digital mobile devices for listening to music nowadays contain thousands of musical titles. Browsing music collections is becoming more and more difficult and time-consuming, and many outdoors listening situations do not allow the user to distract the attention from the current activity and spend too much time in finding the song that fits best with the current mood and context. With Music Mood Wheel, We propose a simple audio interaction model and prototype, in order to make the selection of the desired music contents on a mobile device more comfortable and versatile in the everyday life.
 
We started from the idea that a stimulus for listening to music rarely comes up with a concrete name or song title: but rather with a wish to create a specific ambient, to reinforce a pleasant emotion, to fill up a space or to suppress boredom. Besides, the size of private music collections overwhelmingly exceeds a person's ability to recall which compositions comply best to the current mood. In our project we challenge the "recall and search by title" procedure exhibited by most digital music mobile devices with the “radio wheel” metaphor: When listening to a “vintage radio device”, we simply turn the frequency wheel around until we hear something that catches the attention. As soon as we lose interest in the contents currently broadcasted, we choose another station, by moving the wheel again. One of the most exciting challenges in this project is to check if the recent advances in extracting high level descriptors from the audio signal, like tonal strength, BPM or mean audio loudness estimation can contribute in achieving our vision.
 
To implement the “Music Mood Wheel” vision we interleave software prototyping and user experiments activities. Taking benefit from audio signal processing techniques provided by the Music Technology Group (MTG) at University Pompeu Fabra (UPF) in Barcelona and from our experimental corpus, we designed a "browsing engine" allowing the user to simply, freely and quickly explore his music collection just following "human sensations" (melody and rhythm). We first developed a "rough" prototype on a Tablet PC running Windows XP, and ran a serie of user experiments with 55 students from the University of Milan, to compare this "new device" to off the shelves digital music players.

The user experiments and tests we did so far encourage us in our approach and intuition: the eyes free, pure audio browsing approach is a promising path to follow. We now use the experimental corpus and refined algorithms for completing our final prototype, targeting devices running Windows Mobile 5.0 (e.g smartphones with hard drive storage).

Links

Computing Music Lab, Università degli Studi di Milano  
Music Technology Group, Pompeu Fabra University, Barcelona
Mobile Music Technology – Third international workshop – University of Sussex, Brighton, UK, 2-3 March 2006
Axmedis 2006 – Second international conference on automated production of cross-media content for multi-channel production – Leeds, UK December 13-15, 2006

Pierre-Louis Xech's home page.