Bandpass filtering EEG data

Hey guys,

I am currently messing around with ways to filter EEG data I have recorded. I am currently trying bandpass filtering my data, but the data after filtering looks funky. The raw data I record is measured in mV, and has mostly "small" values, mainly <0.5 but occasionally >1. After using the bandpass filter below, data look funny, with values become "huge" (y axis goes from -400:400). I am hoping someone can give me some basic pointers on what filter parameters I need to change.

      Fstop1 = 0.1; % First Stopband Frequency
      Fpass1 = 0.2; % First Passband Frequency
      Fpass2 = 95;  % Second Passband Frequency
      Fstop2 = 100; % Second Stopband Frequency
      Astop1 = 60;  % First Stopband Attenuation (dB)
      Apass  = 1;   % Passband Ripple (dB)
      Astop2 = 80;  % Second Stopband Attenuation (dB)
      match  = 'stopband'; % Band to match exactly

      % Construct an FDESIGN object and call its BUTTER method.
      bpfilt = fdesign.bandpass(Fstop1,Fpass1,Fpass2,Fstop2,Astop1,Apass,Astop2,Fs);
      Hd     = design(bpfilt, 'butter', 'MatchExactly', match);

      eegdata = filter(Hd,eegdata);

I have attached images on the data plotted before and after using this filter.

Answer 1

[Star Strider]

I can’t run your code because I don’t have ‘Fs’. I almost always use the second-order-section implementation of a filter, since those tend to be stable when transfer functions are not.

Before you actually use your filter, be sure to look at the filter passband with freqz. This will tell you how stable it is and what the passband actually looks like.

[Pavlov]

Also, can you explain what 'second-order-section implementation' and 'transfer function' mean?

[Star Strider]

Second-order sections are a bit much to explain here, so I’ll refer you to the documentation on tf2sos. I don’t see an automatic option for second-order section implementation in fdesign. I’m not that familiar with it, since I use the lower-level functions to design my filters.

A ‘transfer function’ is the frequency response of a filter (or network, control system, etc.), usually displayed as a Bode plot with amplitude and phase.

Noted for later:

Fs = 4KHz

It’s the end of my day as well.

Answer 2

[Pavlov]

Thanks, I will try this tomorrow as I am heading home for the night. Fs = 4000 in my case.