Announcement

Collapse
No announcement yet.

Why do I do a lowpass in the end of the processing of EMG?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Why do I do a lowpass in the end of the processing of EMG?

    I am using a Matlab code to process the data of EMG during a gait. The code is Below:

    raw is the raw EMG, collected with 1500Hz
    bandpass=[20,450]; % Bandpas filter
    lowpass= 10; % Low pass filter: smooth data
    % 1. remove the offset
    raw_DC=raw-mean(raw);
    % 2. bandpass filter: butterworth, 4th order, 20-450Hz
    [b,a] = butter(4, bandpass/750);
    raw_band=filtfilt(b,a,raw_DC);
    % 3. rectify
    raw_rec= abs(raw_band);
    % 4. smooth by lowpass filter: 4th order, 10 Hz
    [b, a] = butter(4,lowpass/750,'low');
    raw_low=filtfilt(b,a,raw_rec);

    My question is: I have filtered the EMG with the bandpass filter of 20-450Hz. Does it mean there is no EMG with the frequency below 20Hz? So why I should filter the rectified EMG with lowpass of 10Hz in the end? Or how can I do that, because there is no EMG below 20Hz after the bandpass as I assumed.

    Thanks
    Last edited by Michael Han; April 26, 2013, 05:22 PM.

  • #2
    Re: Why do I do a lowpass in the end of the processing of EMG?

    After filtering and rectification, the rectified EMG data will look like this:




    But most likely what you are expecting is a smoother envelope like this:



    Filtering the rectified EMG data is done to smooth the data to generate the envelope above.

    Comment


    • #3
      Re: Why do I do a lowpass in the end of the processing of EMG?

      Originally posted by ecramp48 View Post
      After filtering and rectification, the rectified EMG data will look like this:




      But most likely what you are expecting is a smoother envelope like this:



      Filtering the rectified EMG data is done to smooth the data to generate the envelope above.
      --------------------------------------------------------------------------------
      Thank you. I have got the same result. My question is after the bandpass between [20 450] Hz, my comprehension is that the frequency below 20 Hz has been filtered. So why the signal can still be filtered by a 10 Hz lowpass? I think I have some misunderstanding about the filter of the signal?

      Comment


      • #4
        Re: Why do I do a lowpass in the end of the processing of EMG?

        Hi Michael,

        If you do the bandpass filter then full-wave rectify after that, the rectification step changes the frequency content of the signal. Think about a 1-Hz sine wave then "fold" the negative bumps over: you don't have a simple 1-Hz waveform anymore. So you will end up with frequency content that was not there before you rectified.

        Ross

        Comment


        • #5
          Re: Why do I do a lowpass in the end of the processing of EMG?

          Thanks Ross. I was thinking about the rectifying. after the rectifying, the negative peaks were put in the opposite direction, but it does not mean the frequency of the wave was doubled? it just means that the frequency was changed, maybe doubled, even more? should i understand it in this way? or do u have some suggestion about what book or which chapter I should read about the spectrum analysis, especially about the rectify?

          Comment


          • #6
            Re: Why do I do a lowpass in the end of the processing of EMG?

            Originally posted by Michael View Post
            Thanks Ross. I was thinking about the rectifying. after the rectifying, the negative peaks were put in the opposite direction, but it does not mean the frequency of the wave was doubled? it just means that the frequency was changed, maybe doubled, even more? should i understand it in this way? or do u have some suggestion about what book or which chapter I should read about the spectrum analysis, especially about the rectify?
            Rectifying any signal really kills all the high-frequency content. You can verify this easily by plotting a power spectrum of your EMG prior to and following rectification.

            Cheers,
            -Ian

            Comment


            • #7
              Re: Why do I do a lowpass in the end of the processing of EMG?

              Hi Michael,
              I'm unable to help with your book or chapter inquiry. However, it would be good to note that rectification of sEMG signal is a non-linear process in regards to signal content. So, it can't be said that it predictably "doubles" existing frequency content or eliminates frequency content in any certain bands.

              Matt

              Comment


              • #8
                Re: Why do I do a lowpass in the end of the processing of EMG?

                Originally posted by ikremenic47 View Post
                Rectifying any signal really kills all the high-frequency content. You can verify this easily by plotting a power spectrum of your EMG prior to and following rectification.

                Cheers,
                -Ian
                ---------------------------------------------------------
                Thank you Ian. I did it and found what you said.


                raw_band_spectral_analysis.jpgraw_rectify_spectral_analysis.jpg

                above is the plot.
                I know the rectifying kill the high frequency now. So why can rectifying do that so? That's because the principle of rectifying and the FFT? That's just a nature of the signal here? can I understand it so?

                Comment


                • #9
                  Re: Why do I do a lowpass in the end of the processing of EMG?

                  Thank you Matt.
                  I did the spectrul analysis for the signal before and after rectifying, and found the high frequency was killed by the rectifying as Ian said. I think your answer is right. It is a non-linear process. What I know now is that is the nature of the signal here.

                  Comment


                  • #10
                    Re: Why do I do a lowpass in the end of the processing of EMG?

                    A further point is that the final "envelope" is supposed to approximate the muscle tension. In effect, the excitation-contraction coupling and muscle fiber properties (series elastic, parallel elastic etc.) behave as a low-pass filter on the EMG input, such that the final force output is much smoother. Ideally, the cut-off of the final low-pass filter should be set to the twitch time of the muscle, with slow muscles such as quads having a lower cut-off than faster muscles, such as hamstrings. In practice, I don't think anyone bothers to do this - I guess it won't make much difference, but at least you can now see the original idea of the low-pass filter envelope.

                    Comment


                    • #11
                      Re: Why do I do a lowpass in the end of the processing of EMG?

                      thanks Chris

                      Comment

                      Working...
                      X