Biomedical Engineering Reference
In-Depth Information
sia in volunteers [35]. While simple to calculate in the era before inexpensive com-
puter chips, the ZXF parameter is not simply related to frequency-domain estimates
of frequency content as demonstrated in Figure 9.3, because not all frequency com-
ponent waves in the signal will cross the zero point.
Demetrescu refined the zero crossing concept to produce what he termed
aperiodic analysis
[36]. This method simply splits the EEG into two frequency bands
(0.5 to 7.9 and 8 to 29.9 Hz) and the filtered waveforms from the high-and low-fre-
quency bands are each separately sent to a relative minima detector. Here, a wavelet
is defined as a voltage fluctuation between adjacent minima, and its frequency is
defined as the reciprocal of the time between the waves. Wavelet amplitude is
defined as the difference between the intervening maxima and the average of the two
minima voltages. Aperiodic analysis produces a spectrum-like display which plots a
sampling of detected wavelets as an array of “telephone poles” whose height repre-
sents measured wave amplitude, distance from the left edge frequency (in a logarith-
mic scale), and distance from the lower edge time since occurrence. The Lifescan
Monitor (Diatek, San Diego, California) was an implementation of aperiodic analy-
sis; it is not commercially available at present but the algorithms are described in
detail in paper by Gregory and Pettus [37]. Reports in the literature have used this
Zero crossing frequency and its limitations
EEG
μ
Volts
Time
T1
T2
T3
T4
(a)
EEG
μ
Volts
Time
T1
T2
T3
T4
(b)
Figure 9.3
(a, b) Failure of zero crossing algorithm to be sensitive to all components of EEG wave-
form. In interval T4 and beyond, the high-frequency, low-amplitude activity in waveform B is
ignored.
Search WWH ::
Custom Search