Why do we need process sEMG?
: Raw sEMG signal has many artifact(Motion, ECG e.t.c.). And usage of sEMG is diffrent.
So we have to process Raw sEMG signals.
There are so many artifact of sEMG signals. But I introduce mainly occured artifact of sEMG signals.
1. ECG noise
: some part of muscle near vein can be mixed with sEMG signal -> no solution for this now…
<EMG with ECG noise>
2. Muscle crosstalk
: Two more muscle make sEMG signal at the sametime -> cross correlation of two signals
3. AC interference
: Wire noise and 60Hz noise -> Use 59.5~60.6 notch filter and twist cable.
<EMG AC interference noise>
4. DC – offset
: DC bias is not matched because of skin artifact -> Use high pass filter(5~10Hz)
5. Motion artifact
: Sensor is struck from skin by moving -> Prevent electrodes from being struck
<EMG with motion artifact(Center of signal is come from motion artifact)>
6. Amplifier saturation
: sEMG signals is much bigger amplitude then amplifier -> Larger range of amplifier and A/D converter
After removing artifact, we are doing postprocessing for usage. Here are some technic of sEMG postprocessing.
1. Moving Average
System Architecture : Slide window -> Average
1) Slide window : 100~200ms size and moved over the length of the sample.
2) Average : Make window average.
Very commonly used technic to preprocessing sEMG signals.
But result can be changed by window size.
<Moving Average EMG (Green:Raw EMG, Black : Averaged EMG)>
System Architecture : Slide window -> RMS
1) Slide window : same as moving average’s slide window
2) RMS : Make window RMS
Not using average, so much bigger amplitude than averge. Also has problem of window size.
<Raw EMG and RMS EMG>
3. Integration EMG(IEMG)
Architecture : Full rectification -> Integration
1) Full rectification : Rectified input sEMG signals.
2) Integration : Integrated by user defined time gap.
Time gap can be problem.
<IEMG (Green: IEMG, Red : Raw EMG)>
4. Linear Envelopement
System Architecture : Full Rectification -> Low pass filter
1) Full rectification : same as IEMG’s Full rectification
2) Low pass filter : filtering high frequency make smooth signal.(under 10Hz)
for Slow motion and, hard to know artifact.
<Linear Envelopement Method>
You can choose any of these postprocessing method. But it is not absolute answer. To days many researcheres
developed postprocessing of sEMG signals such as Ensembled method, wavelet methods etc.