Hi,
Thank you very much, Eric, for the response; we greatly appreciate it. So I'd like to show what we're putting into tfrscalo and what we're getting out.
This is our code:
tfrscalo(
signal,
1:length(signal),
wavenumber [we used:
1],
lower_frequency_bound [we used
0.030 to indicate 30 Hz at sampling rate of 1000 Hz] ,
upper_frequency_bound [we used
0.100 to indicate 100 Hz at sampling rate of 1000 Hz] ,
number_of_frequency_bins [we used
14, so that 70Hz/14 = 5 Hz per frequency bin],
0);
We took the first parameter of the 1-by-4 output matrix. That first parameter, to our understanding, is the modulus squared of the complex number you've referred to. Our wavelet output parameters were in the range of negative 1.1 million to positive 9 million.
The 'signal' that is fed into the tfrscalo is a sequence of 1200 separate values, representing the amplitudes of brain waves over 1.2 seconds at a sampling rate of 1000 Hz. Amplitudes can be positive and negative.
Quick restatement of questions:
1) Why are we getting negative values?
2) If we feed tfrscalo a time( seconds) - amplitude (volts) signal, what are
the units of the values in the output matrix?
Thanks a lot,
Yasemin
On Fri, Oct 10, 2008 at 6:33 AM, Eric Chassande-Mottin
<address@hidden> wrote:
dear Yasemin,
the output matrix of tfrscalo should not have negative values
as it is a modulus square of a complex quantity. this is indeed
what i observe when computing
sig=altes(64,0.1,0.45); tfr=tfrscalo(sig); any(tfr(:)<0)
ans = 0
(this is taken from the example given in the online help).
it is likely that you are not using tfrscalo properly.
if you want to get further help, the easiest would be
that you post to this list a small example (like the above) that
reproduce the problem you have.
cheers,
eric
On Thu, 9 Oct 2008, Yasemin Gokce wrote:
Dear Ladies and Gentleman,
I have a question on the use of tfrscalo use. We are analyzing brain waves
and we would like to use tfrscalo to compute the spectrogram (frequency vs
time) from a time-amplitude signal.
Our high-level understanding of the process is that we should be feeding the
time-amplitude signal into tfrscalo. The out put is a 1-by-4 matrix. And the
first value in the output is the magnitude squared of the wavelet transform.
To normalize the values of the output matrix, we've been informed that we
should scale down our values by a certain constant and then take the square
root of the data. Apparently, for other researchers, this method works and
they get a reasonable range that you might expect from brain wave data.
However, we cannot take the square root of the values in our tfrscalo output
matrix, because some values are negative. I was wondering
1) Is it possible for the output matrix of tfrscalo to have negative values?
2) If we feed tfrscalo a time( seconds) - amplitude (volts) signal, what are
the units of the values in the output matrix?
Best regards,
Yasemin