Prev: mpeg heade4r
Next: Chebyshev fitler
From: Vivek Subramanian on 29 Jul 2010 07:22
Hi Everyone,

I'm having some issues using a Chebyshev filter for EEG data which I am extracting. For some reason, there is an unusual sinc function superimposed on my signal after I apply the filter. Here is the process I'm following:

1. Extract raw EEG data: http://uploading.com/files/5641c89e/RawData.fig/

2. Filter the data using a Chebyshev filter: http://uploading.com/files/1ed9emd7/AfterFiltering.fig/
Here is the code for initialization of the filter:
wp =[47 54]/(Fs/2);
ws =[48.5 52.5]/(Fs/2);
[n, wn]=cheb1ord(wp,ws,3,30);
[b, a] = cheby1(n,3,wn,'stop');

3. Continue signal processing and feature extraction.

Could someone please explain the reason for the unusual oscillations present at the beginning of the filtered EEG? Just FYI, my EEG signal is just noise because it takes time to prepare for an EEG, so I prefer to just have my machine acquire noise until I am ready to do an actual experiment.

Thank you so much for your help!!
From: Wayne King on 29 Jul 2010 08:06
"Vivek Subramanian" <vas11(a)duke.edu> wrote in message <i2ro8t$ihv$1(a)fred.mathworks.com>...
> Hi Everyone,
>
> I'm having some issues using a Chebyshev filter for EEG data which I am extracting. For some reason, there is an unusual sinc function superimposed on my signal after I apply the filter. Here is the process I'm following:
>
> 1. Extract raw EEG data: http://uploading.com/files/5641c89e/RawData.fig/
>
> 2. Filter the data using a Chebyshev filter: http://uploading.com/files/1ed9emd7/AfterFiltering.fig/
> Here is the code for initialization of the filter:
> wp =[47 54]/(Fs/2);
> ws =[48.5 52.5]/(Fs/2);
> [n, wn]=cheb1ord(wp,ws,3,30);
> [b, a] = cheby1(n,3,wn,'stop');
>
> 3. Continue signal processing and feature extraction.
>
> Could someone please explain the reason for the unusual oscillations present at the beginning of the filtered EEG? Just FYI, my EEG signal is just noise because it takes time to prepare for an EEG, so I prefer to just have my machine acquire noise until I am ready to do an actual experiment.
>
> Thank you so much for your help!!

Hi Vivek, I'm having difficulty viewing those images. Perhaps you can give us the relevant details so somebody can simulate the behavior. For example, you have not given us the sampling frequency. You give it only has a variable, Fs, but we don't know the value.

Also, what is the approximate bandwidth of the EEG data? I assume that the acquisition process itself involves bandpass filtering.

Wayne
From: Vivek Subramanian on 29 Jul 2010 12:04
Hi Wayne,

Thanks for your reply. Sorry, I chose a bad site to upload files. Perhaps this will be better:

Raw EEG data: http://www.mediafire.com/file/byj3qx35urtbpri/RawData.fig

After Chebyshev filter: http://www.mediafire.com/file/juehjea15ylvilj/AfterFiltering.fig

My Fs is 128. The bandwidth of the EEG ranges from DC to 60 Hz. I, however, only want the first 45 frequencies (i.e. 1 Hz, 2 Hz, 3 Hz,..., 45 Hz), encapsulating the alpha, beta, and gamma bands. Theoretically, the 50 Hz component should not affect the rest of the signal if I use a low-pass filter (since 45 < 50), but it is so strong, that nearby frequencies components are being affected by it. This is why I'm trying to first design a bandstop (notch) filter before extracting the first 45 frequency components.

Please let me know what you think!
Vivek

"Wayne King" <wmkingty(a)gmail.com> wrote in message <i2rqrd$5jk$1(a)fred.mathworks.com>...
> "Vivek Subramanian" <vas11(a)duke.edu> wrote in message <i2ro8t$ihv$1(a)fred.mathworks.com>...
> > Hi Everyone,
> >
> > I'm having some issues using a Chebyshev filter for EEG data which I am extracting. For some reason, there is an unusual sinc function superimposed on my signal after I apply the filter. Here is the process I'm following:
> >
> > 1. Extract raw EEG data: http://uploading.com/files/5641c89e/RawData.fig/
> >
> > 2. Filter the data using a Chebyshev filter: http://uploading.com/files/1ed9emd7/AfterFiltering.fig/
> > Here is the code for initialization of the filter:
> > wp =[47 54]/(Fs/2);
> > ws =[48.5 52.5]/(Fs/2);
> > [n, wn]=cheb1ord(wp,ws,3,30);
> > [b, a] = cheby1(n,3,wn,'stop');
> >
> > 3. Continue signal processing and feature extraction.
> >
> > Could someone please explain the reason for the unusual oscillations present at the beginning of the filtered EEG? Just FYI, my EEG signal is just noise because it takes time to prepare for an EEG, so I prefer to just have my machine acquire noise until I am ready to do an actual experiment.
> >
> > Thank you so much for your help!!
>
> Hi Vivek, I'm having difficulty viewing those images. Perhaps you can give us the relevant details so somebody can simulate the behavior. For example, you have not given us the sampling frequency. You give it only has a variable, Fs, but we don't know the value.
>
> Also, what is the approximate bandwidth of the EEG data? I assume that the acquisition process itself involves bandpass filtering.
>
> Wayne
From: Vivek Subramanian on 31 Jul 2010 08:49
"Vivek Subramanian" <vas11(a)duke.edu> wrote in message <i2s8pk$nnm$1(a)fred.mathworks.com>...
> Hi Wayne,
>
> Thanks for your reply. Sorry, I chose a bad site to upload files. Perhaps this will be better:
>
> Raw EEG data: http://www.mediafire.com/file/byj3qx35urtbpri/RawData.fig
>
> After Chebyshev filter: http://www.mediafire.com/file/juehjea15ylvilj/AfterFiltering.fig
>
> My Fs is 128. The bandwidth of the EEG ranges from DC to 60 Hz. I, however, only want the first 45 frequencies (i.e. 1 Hz, 2 Hz, 3 Hz,..., 45 Hz), encapsulating the alpha, beta, and gamma bands. Theoretically, the 50 Hz component should not affect the rest of the signal if I use a low-pass filter (since 45 < 50), but it is so strong, that nearby frequencies components are being affected by it. This is why I'm trying to first design a bandstop (notch) filter before extracting the first 45 frequency components.
>
> Please let me know what you think!
> Vivek
>
> "Wayne King" <wmkingty(a)gmail.com> wrote in message <i2rqrd$5jk$1(a)fred.mathworks.com>...
> > "Vivek Subramanian" <vas11(a)duke.edu> wrote in message <i2ro8t$ihv$1(a)fred.mathworks.com>...
> > > Hi Everyone,
> > >
> > > I'm having some issues using a Chebyshev filter for EEG data which I am extracting. For some reason, there is an unusual sinc function superimposed on my signal after I apply the filter. Here is the process I'm following:
> > >
> > > 1. Extract raw EEG data: http://uploading.com/files/5641c89e/RawData.fig/
> > >
> > > 2. Filter the data using a Chebyshev filter: http://uploading.com/files/1ed9emd7/AfterFiltering.fig/
> > > Here is the code for initialization of the filter:
> > > wp =[47 54]/(Fs/2);
> > > ws =[48.5 52.5]/(Fs/2);
> > > [n, wn]=cheb1ord(wp,ws,3,30);
> > > [b, a] = cheby1(n,3,wn,'stop');
> > >
> > > 3. Continue signal processing and feature extraction.
> > >
> > > Could someone please explain the reason for the unusual oscillations present at the beginning of the filtered EEG? Just FYI, my EEG signal is just noise because it takes time to prepare for an EEG, so I prefer to just have my machine acquire noise until I am ready to do an actual experiment.
> > >
> > > Thank you so much for your help!!
> >
> > Hi Vivek, I'm having difficulty viewing those images. Perhaps you can give us the relevant details so somebody can simulate the behavior. For example, you have not given us the sampling frequency. You give it only has a variable, Fs, but we don't know the value.
> >
> > Also, what is the approximate bandwidth of the EEG data? I assume that the acquisition process itself involves bandpass filtering.
> >
> > Wayne

Hi Everyone,

I actually figured out what was going wrong. The 1D filter was causing the sinusoids to show up, but when I used a forward and reverse filter (filtfilt), the problem was solved.

Thanks for your help!
Vivek
 | 
Pages: 1
Prev: mpeg heade4r
Next: Chebyshev fitler