Biomedical User Group Discussions

cancel
Showing results for 
Search instead for 
Did you mean: 

EMG Signal Linear Regression implementation

Hi Michel,

could I ask you to validate my mean power frequency linear regression implementation, please?

Here is the vi and a acquired emg.

I reduce the window from 0,5 to 0,25 ms for better resolution.

Thank you!

Ramon Campos

Download All
0 Kudos
Message 21 of 26
(1,218 Views)

Hi Ramon,

    I'm still using LabVIEW 2009 here so I can't open your vi.

Salutations,

Michel

0 Kudos
Message 22 of 26
(1,218 Views)

I'm sorry!!

here for the lv 2009.

Ramon Campos

0 Kudos
Message 23 of 26
(1,218 Views)

Hi Ramon,

    I don't have the PDS-FFT vi but if the output is half the spectrum it's ok.  You just need to change the loop of the "mean power frequency".  You need to multiply each of the spectrum amplitude value by it's associated frequency (and not the index of the loop).  Here's the image of the changed loop:

Ramon_MPF.JPG

As for the time axis, I guess this is correct if the correct time is associated with the correct data block.  Hope this helps.

Addenda: You could also use the loop as you did and multiply the end result (after dividing by PSUM) by "df" (second output of the cluster).  That would give the same result.

Salutations,

Michel

0 Kudos
Message 24 of 26
(1,218 Views)

Hello Michel.

I want to thank you one more time for the help.

Here some results from my implementation:

FFT in some instante of time:

FFT.png

MDF - linear slope: -0,260339

MDF.png

MPF - linear slope: -0,38142

MPF.png

i'm trying to implement some wavelet analysis too.

I'll get wavelet coefficients (with scales 4, 8, 16, 32, 64, 128) and do the RMS linear regression.

An other analysis i'll get the frequency from the wavelet scalogram and apply mdf and mpf linear regression.

When I get some results I'll post here.

Abraços!

Ramon Campos

0 Kudos
Message 25 of 26
(1,218 Views)

Hi Ramon,

    Nice work!

    In your FFT, you seem to have some line noise (and maybe other) at 50 Hz (i guess this is the sector frequency where you plug your devices) that reflects as a peak of frequency.  You should use some method to attenuate this artefact before computing your MPF/MDF.  Unfortunately, some signal information exists around this frequency so you have to watch-out to not remove too much of the "real" signal.

Salutations,

Michel

0 Kudos
Message 26 of 26
(1,218 Views)