Hi all,
I am currently trying to carry out an EMG coherence analysis, prior to this I have been reading some papers.
I can understand the Bandpass implication from range A Hz to B Hz, but from Boonstra's paper from 2012, they also apply high pass filter at 250Hz, which I cannot quite get my head around it.
If we apply high pass filter at 250Hz, do we not lose all the data below 250Hz? but the paper says that it actually increased the coherence. Could anyone care to explain this?
Spectral analysis from my own data is ran, and I still don't quite understand how applying high-pass filter after band-pass then rectifying improves the coherence?
Hi Yong Kuk Kim
For coherence between muscles, what you are really trying to do is identify simultaneous/phase-locked patterns in the timing or density of motor unit action potentials. That's not the same as looking for phase-locked continuous oscillations... think 'point process' not 'waves' here. Accordingly, high pass filtering is meant to remove anything that isn't a motor unit action potential, basically turning the surface EMG signals into something closer to multi-unit spike trains. Even the shapes of the action potentials have some spectral content which is not meaningful and is best to remove.
I've had good luck using 250 Hz at low-medium forces using small high-quality electrodes. It's possible that 250 may be too high for other muscles/setups where the EMG signals are not as spiky to begin with. I don't think there would ever be a good reason to go under 40Hz though.
Regarding rectification, or any method of amplitude envelope extraction, it helps identify coherent fluctuations in the density of action potentials over time and/or periodic timing of simultaneous action potentials. Rectification may be necessary when looking for very low frequencies of coherence, especially at lower forces where there are fewer action potentials overall, but at very high forces it may be detrimental. In that case, too many action potential shapes collide at the electrode and decrease/smooth the EMG amplitude envelope.
Unfortunately, for typical surface EMG, 'optimal' processing for coherence analysis may take some trial and error - but usually high pass filtering followed by rectification is a good place to start.
- Badges
- Science topic
- Similar topics
- Control Systems Engineering
- Filtering