Music and Human Activity

Download Report

Transcript Music and Human Activity

Jesse English and
Sandor Dornbush
With help from:
Dr. Zary Segall &
Chad Eby.
Activity Awareness



Feedback loop
between the user and
their music.
Use the users
physiological state to
influence the music
selection.
The music influences
the users mood and
mindset.
Proposed XPod Form Factor
Proposed XPod User Interface
Potential XPod Platform



Nokia 5500 Sport
Phone
Embedded
Accelerometers
MP3 Player
http://europe.nokia.com/A4213113
Experimental XPod
•Bodymedia Device
•Tablet
Music and Human Activity



Music has a strong connection with human activity.
People like to play music that is appropriate for their
current activity.
Music will can affect a persons mental state and
physical activity.
Music Selection



Choose the song that the user would most likely
like in this state.
Use the user prior behavior to learn their
preferences.
Combine a star rating system with body
information.
Song Selection
Choose a
Song Randomly
Learning Algorithm
Estimate Preference
Flip coin weighted by the
Estimated preference
Heads
Play Song
Tails
Results
Percent Correct
55%
45%
35%
25%
J48
Without State
AdaBoost
With State
SMO
IBk
Algorithm
Neural Net
Results


XPod creates a custom context aware playlist.
The addition of state information can improve
the accuracy of the learning system.
–

Neural Network seems to be best.
–
–

Prefrence(Song|Time,Activity) ≠ Preference(Song)
Generalizes best
Over trains very quickly
SMO gets the exact.
–
Does not generalize well.


289 Input neurons
1 Output neuron
1-50 Hidden neurons
–
4 is the best
50%
45%
40%
35%
30%
25%
1
2
3
4
Without State
5
10 15 20 25 30 35 40 45 50
With State
Neurons in Hidden Layer
0.2
0.195
RMSE

% Accurately Classified
Network Structure
0.19
0.185
0.18
0.175
0.17
1
2
Without State
3
4
5
10 15 20 25 30 35 40 45 50
With State
Neurons in Hidden Layer
Future Work

Develop a prototype device.
–
–

Incorporate song metadata
–
–

Nokia 5500 Sport
iPaq and Backpaq
Human Generated, eg Last.fm
Machine Generated
Incorporate other meta information
–
–
–
Location Information
Recent Phone Calls
Weather
Questions?