EEG and Brain Connectivity: A Tutorial - Bio-Medical Instruments, Inc.
EEG and Brain Connectivity: A Tutorial - Bio-Medical Instruments, Inc.
EEG and Brain Connectivity: A Tutorial - Bio-Medical Instruments, Inc.
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
The time series can be divided into two general categories: 1- the statistics of short-time<br />
biological data <strong>and</strong> other short-time interval events such as economic, sociological, etc.<br />
<strong>and</strong> 2- long time span events such as astronomical, meterological, geologic <strong>and</strong><br />
geophysical data . . . . . . . – to be continued<br />
44- Appendix – B<br />
Instantaneous Coherence <strong>and</strong> Phase Difference<br />
Complex demodulation was used to compute instantaneous coherence <strong>and</strong> phasedifferences<br />
(Granger <strong>and</strong> Hatanaka, 1964; Otnes <strong>and</strong> Enochson, 1972; Bloomfield, 2000).<br />
This method first multiples a time series by the complex function of a sine <strong>and</strong> cosine at a<br />
particular frequency followed by a low pass filter which removes all but very low<br />
frequencies <strong>and</strong> transforms the time series into instantaneous amplitude <strong>and</strong> phase <strong>and</strong> an<br />
“instantaneous” spectrum (Bloomfield, 2000). We place quotations around the term<br />
“instantaneous” to emphasize that there is always a trade-off between time resolution <strong>and</strong><br />
frequency resolution. The broader the b<strong>and</strong> width the higher the time resolution but the<br />
lower the frequency resolution <strong>and</strong> vice versa (Bloomfield, 2000). For example, if we<br />
multiply a time series {x t , t = 1, . . . , n} by sine ω 0 t <strong>and</strong> cos ω 0 t <strong>and</strong> then apply a low pass<br />
filter F, we have<br />
Z′<br />
t<br />
= F( xt<br />
sinω<br />
0t)<br />
Z′<br />
′= F( x cosω<br />
0t)<br />
t<br />
2<br />
[ ] 1 / 2<br />
2<br />
<strong>and</strong> ( Z ) + ( Z′<br />
)<br />
t<br />
2 ′<br />
t<br />
t<br />
is an estimate of the “instantaneous” amplitude of the<br />
Z′<br />
−1 t<br />
frequency ω 0 at time t <strong>and</strong> tan is an estimate of the “instantaneous” phase at<br />
′′<br />
Z<br />
t<br />
time t.<br />
The instantaneous cross-spectrum is computed when there are two time series {y t ,<br />
t = 1, . . . , n} <strong>and</strong> {y’ t , t = 1, . . . , n} <strong>and</strong> if F [ ] is a filter passing only frequencies near<br />
zero, then, as above<br />
R<br />
t<br />
2<br />
2<br />
2<br />
iω<br />
[ ] [ ] [ ] 2 0t<br />
y sinω t + F y cosω<br />
t = F y e<br />
= F<br />
is the estimate of the<br />
amplitude of frequency ω 0 at time t <strong>and</strong><br />
t<br />
0<br />
[ y ] ⎞<br />
t<br />
sinω<br />
0t<br />
[ y cosω<br />
t] ⎟⎟<br />
t<br />
−1⎛<br />
F<br />
ϕ<br />
t<br />
= tan ⎜⎜<br />
is an estimate of the phase of frequency ω 0 at<br />
⎝ F<br />
t 0 ⎠<br />
time t <strong>and</strong> since<br />
0<br />
t