[Eeglablist] time frequency decomposition

Arnaud Delorme arno at salk.edu
Fri Aug 5 16:40:48 PDT 2005


Dear Rachel,

for answering your question, look at the 4 slides generated for one of 
the EEGLAB workshop below. Basically a factor equal 1 indicates real 
wavelet (same number of cycle at all frequency). A factor of 0 indicates 
a standard fourier transform (same window size at all frequencies). A 
factor of 0.5 indicates that the window size at the highest frequency 
will be half the window size at the lowest frequency. The wavelet factor 
for the modified wavelet in timef() are graphicaly explaned in

http://sccn.ucsd.edu/eeglab/workshop2004/Material/Arnaud/Time_frequency_presentation/pwelch_timef_params.pdf

Using this factor is artificial and not optimal. In a next revision for 
these functions, it will actually be possible to determine the actual 
number of cycles at the lowest and at the highest frequency,

Best,

Arno

Rachel Seabrook wrote:

>In the timef command, the cycles parameter gives the option to choose FFT,
>wavelets or multitaper and if wavelets, how many cycles to each time
>window and whether/how this should be varied with frequency.
>
>Firstly, a technical question. If I use the default settings of [3 0.5], I
>understand that this means 3 cycles per window for the lowest frequency
>(though I don't know what the lowest frequency is, or how that is
>determined) but I don't know exactly what the 0.5 means.
>
>Secondly, relating more to the underlying mathematical theory, can anyone
>advise me on how to choose between these options, or point me towards
>references that would help me understand the implications of each? Our
>results depend heavily on which parameters we use - peaks of gamma
>activity come and go with different variations.
>
>Thank you very much,
>
>Rachel
>  
>




More information about the eeglablist mailing list