MATLAB Central Newsreader - Question on filtering

Question on filtering

Florian — Fri, 02 May 2008 09:41:03 -0400

Hi,
I have a electrophysiological signal (electrogastrography).
It is a one channel recording with a main frequency content
in the band between 0 and 0.2Hz, so very low frequencies.
Playing arround with the signal processing toolbox I was
simply overwhelmed by the amount of possible filters.
What I want to do is to apply bandpass filtering and
downsample the signal from 500 Hz to about 1 Hz
(afterwards). The band I am interested in is from 0.01 to
0.2 Hz. Could anybody please tell me a good way to do this?

Thank you very much in advance.

Florian Beissner

Re: Question on filtering

NZTideMan — Sun, 04 May 2008 10:29:27 -0400

On May 2, 9:41=A0pm, "Florian " <beiss...@med.uni-frankfurt.de> wrote:
> Hi,
> I have a electrophysiological signal (electrogastrography).
> It is a one channel recording with a main frequency content
> in the band between 0 and 0.2Hz, so very low frequencies.
> Playing arround with the signal processing toolbox I was
> simply overwhelmed by the amount of possible filters.
> What I want to do is to apply bandpass filtering and
> downsample the signal from 500 Hz to about 1 Hz
> (afterwards). The band I am interested in is from 0.01 to
> 0.2 Hz. Could anybody please tell me a good way to do this?
>
> Thank you very much in advance.
>
> Florian Beissner

downsampling aka decimation
Lots of ways to do it............
Simplest is to just take every 500 th sample, but you may want to
filter first to avoid aliasing.
Simplest filter is the mean of each 500 samples.

Re: Question on filtering

Lars — Sun, 04 May 2008 12:35:03 -0400

I would create a simple filter by using

>> fdatool

say simple IIR-butterworth. you can export the filter from
the tool as numerator (a) and denominator (b) or as a
filter-object.

then you can filter your signal by computing

>> filter(b,a,signal) or >> filter(obj, signal)

when possible, it always is a good idea to keep
computational effort low. --> so perhaps you first do the
downsampling and then you filter the the shorter signal.
for downsampling you can find good ways in the help or you
write your own routine using i.e. a simple fft algorithm.

Regards, Lars

Re: Question on filtering

Lars — Sun, 04 May 2008 12:43:04 -0400

sorry bad mistake:

numerator (b)
denominator (a)

then you can filter your signal by computing

>> filter(b,a,signal)

Regards, Lars

Re: Question on filtering

Bas — Sun, 04 May 2008 17:14:14 -0400

On May 2, 11:41=A0am, "Florian " <beiss...@med.uni-frankfurt.de> wrote:
> Hi,
> I have a electrophysiological signal (electrogastrography).
> It is a one channel recording with a main frequency content
> in the band between 0 and 0.2Hz, so very low frequencies.
> Playing arround with the signal processing toolbox I was
> simply overwhelmed by the amount of possible filters.
> What I want to do is to apply bandpass filtering and
> downsample the signal from 500 Hz to about 1 Hz
> (afterwards). The band I am interested in is from 0.01 to
> 0.2 Hz. Could anybody please tell me a good way to do this?

If you don't care so much about the exact filtering and just want to
be sure you don't have any aliasing, you could try to use 'decimate',
which is basically what you want. According to the docs, it defaults
to a low-pass filter at 0.8 times the nyquist frequency (but have a
look a the the various options). This would be 0.4 Hz in your case,
which seems to match your requirements. There is a warning to do it in
several steps if you want to decimate by more than a factor of 13, so
that would mean 3 steps in your case:
x_filt =3D decimate(decimate(decimate(x, 10), 10), 5);
I only used this function to decimation with small factors, so use at
your own risk.

HTH,
Bas