Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 121

FIGURE 4.9: EEG transmission in awake state eyes closed obtained by means of
different methods.
of DTF. On the other hand making the transformation to the source space involves
calculations, which can change the delicate balance in phase dependencies between
channels. Determination of MVAR model coefficients (and hence DTF) is based on
the correlation matrix between signals. The transformation changes this correlation
matrix in such a way that the correlations may be disturbed. It is recommended to
use as input to DTF the signals referenced to “neutral” electrode, e.g., “linked ears,”
“nose,” etc. Before fitting the MVAR model mean value of the signal should be sub-
tracted, and the data should be normalized dividing by the variance. Preprocessing
influencing the phase relation between channels (e.g., artifact subtraction) should be
avoided. The PDC method similar to DTF is robust in respect to noise and insen-