Information Technology Reference
In-Depth Information
After a few trials, we decided to use seven channels of EEG, which can be
obtained with seven pairs of electrodes placed on the scalp, covering a broad area of
the head. The signals were
filtered in order to tear out signal interference (e.g.
interference generated on electrodes near the eyes due to eye blinking) and added
their signals together prior to executing the analyses. The system analyses the
spectrum of the EEG and its complexity. The analyses yield two streams of control
parameters for the generative music system: one, which carries information about
the most prominent frequency band in the signal
—
popularly referred to as EEG
rhythms
and another, which carries a measure of the complexity of the signal. The
former was used to control algorithms that generated the music, and the latter was
used to regulate the tempo and the loudness of the music.
The most prominent EEG frequency band is obtained with a standard fast
Fourier transform (FFT) algorithm, and the measure of complexity is obtained with
Hjorth
—
s analysis (Hjorth
1970
).
FFT analysis is well known in BCI research and will be discussed in more detail
in other chapters of this volume. Basically, the system looks for two patterns of
information in the spectrum of the EEG: alpha and beta rhythms. Alpha rhythms are
strong frequency components in the signal between 8 and 13 Hz and beta rhythms
are strong components between 14 and 33 Hz.
The less familiar Hjorth
'
is analysis is a time-based amplitude analysis, which
yields three measurements: activity, mobility and complexity. The signal is mea-
sured for successive epochs
'
—
for windows
—
of one to several seconds. Activity and
mobility are obtained from the
first and second time derivatives of amplitude
fluctuations in the signal. The
first derivative is the rate of change of the signal
'
s
amplitude. At peaks and troughs, the
first derivative is zero. At other points, it will
be positive or negative depending on whether the amplitude is increasing or
decreasing with time. The steeper the slope of the wave, the greater will be the
amplitude of the
first derivative. The second derivative is determined by taking the
rst
derivative, which correspond to points of greatest slope in the original signal, result
in zero amplitude in the second derivative and so forth.
Amplitude
first derivative of the
first derivative of the signal. Peaks and troughs in the
fluctuations in the epoch give a measure of activity. Mobility is
calculated by taking the square root of the variance of the
first derivative divided by
the variance of the primary signal. Complexity is the ratio of the mobility of the
rst
derivative of the signal to the mobility of the signal itself; for instance, a sine wave
has a complexity equal to 1. Figure
1.2
shows an example of Hjorth analysis. A raw
EEG signal is plotted at the top (C:1), and its respective Hjorth analysis is plotted
below: activity (C:2), mobility (C:3) and complexity (C:4). The tempo of the music
is modulated by the complexity measure.
BCMI-Piano
s music algorithm was developed with the assistance of Bram
Boskamp, then a postgraduate student at ICCMR. It generates the music using rules
that are deduced automatically from a given corpus of examples. It deduces
sequencing rules and creates a transition matrix representing the transition logic of
what follows what. New musical pieces are generated in the style of the ones of the
training corpus. Firstly, the system extracts blocks of music and deduces the rules
'
