Bonjour,
Je suis post-doctorant en neurosciences et il y très souvent fait référence dans les publications à une analyse de cohérence entre deux signaux (par exemple entre les décharges d'un neurone et le champ magnètique local). La cohérence est définit ici:
Rate histograms are used to estimate the power spectra of the spike trains. The
p arameters of rate histograms are calculated using the following formulas:
B in = 1./(2.* Maximum_Frequency )
N umber_of_Bins = 2* Number_of_Frequency_Values
For two variables X (reference) and Y (target) the following entities are calculated. The time intervals
of length Bin*NumberOfBins with specified overlaps are defined. For each interval, rate histograms
are calculated and detrended and their FFTs are calculated. Then, individual and cross-densities are
calculated:
P xx = FFT(X)*Conj(FFT(X)),
P yy = FFT(Y)*Conj(FFT(Y))
P xy = FFT(X)*Conj(FFT(Y)).
Here Conj(z) is a complex conjugate of z. Pxx, Pyy and Pxy values are averaged across all intervals
and coherence values are calculated as
Mean(Pxy)*Mean(Pxy) / (Mean(Pxx)*Mean(Pyy)).
Coherence phase values are calculated as phase of Mean(Pxy).
For continuous variables, if both the reference and the target variables have the same digitizing
frequency, the FFTs of variable values are calculated and then resampled to the specified frequency
steps. If the reference is a timestamped variable or two continuous variables have different digitizing
rates, the values of continuous variables are averaged within the specified Bins and then FFTs of
these averages are calculated.
J'apprécierais énormément si quelqu'un pourrait me dire dans un langage un peu moins mathématique ce que cela signifie...
Merci
