Information Technology Reference
In-Depth Information
Table 1.2
Average classi
cation scores for the active listening experiment
Subject
Classi
cation task
Mean Min.
Max.
Deviation
Con
dence
1
Active
×
passive
±
0.007
0.998
0.979
1.000
0.007
Active
×
counting
±
0.009
0.996
0.979
1.000
0.009
Passive
×
counting
±
0.010
0.994
0.979
1.000
0.010
Active
×
passive
×
counting
±
0.016
0.998
0.958
1.000
0.015
2
Active
×
passive
±
0.010
0.994
0.979
1.000
0.010
Active
×
counting
±
0.032
0.973
0.896
1.000
0.031
Passive
×
counting
±
0.039
0.954
0.896
1.000
0.038
Active
×
passive
×
counting
±
0.024
0.951
0.903
0.986
0.023
3
Active
×
passive
±
0.014
0.973
0.958
1.000
0.014
Active
×
counting
±
0.011
0.992
0.979
1.000
0.011
Passive
×
counting
±
0.014
0.994
0.958
1.000
0.014
Active
×
passive
×
counting
±
0.016
0.985
0.958
1.000
0.015
establish if a subject is actively listening to music, or passively listening to it
without any special mental effort. This notion is supported by a number of reports
on experiments looking into musical imagination (Meister et al.
2004
; Limb and
Braun
2008
; Miranda et al.
2005
; Petsche et al.
1996
).
1.7.1 Towards an Active Listening BCMI
The results from the above experiment encouraged me to look into the possibility of
developing a BCMI whereby the user would be able to affect the music being
generated in real time by focusing attention to speci
c constituents of the music. I
designed a prototype, which produces two tracks of music of the same style of the
music stimuli that was devised for the experiment: it comprises a rhythmic track and
a solo track, which is generated by means of algorithms that transforms a given riff;
they can transpose it, change rhythm, add a note, remove a note, play the riff
backwards and so on.
Firstly, the neural network is trained to recognise when the incoming EEG
corresponds to active or passive listening, as described in the experimental pro-
cedure. Needless to say, the person who controls the music here should be the same
as the one who
s EEG was used to train the system. The system works as follows:
the rhythmic part is continuously played and a riff is played sporadically; an initial
riff is given by default. Immediately after a riff is played, the system checks the
subject
'
s EEG. If it detects active listening behaviour, then the system applies some
transformation on the riff that has just been played and plays it again. Otherwise, it
does not do anything to the riff and waits for the subject
'
'
s EEG response to the next
