Mega and Noraxon Frequency Shifts for Fatigue
Noraxon
Mega
0 10 20 30 40 50 60
0 10 20 30 40 50 60 70 80 90 100
Average Frequency
Time 80.682
11.209 AVF 10 DEVF⋅ REG
56.32
0 TIMEK
0 10 20 30 40 50 60
0 10 20 30 40 50 60 70 80 90 100
Average Frequency
Time 90.112
9.181 AVF 10 DEVF⋅ REG
57.344
0 TIMEK
Method. MH lifted a 10 kg weight with one arm for ca 1 minute (static), measuring EMG from the biceps. The first measurement was done using Noraxon (sampling rate 1000 S/s), then after 30 min pause the measurement was repeated using Mega (sampling rate 2000 S/s) employing the same electrodes.
The frequency shifts were calculated by dividing the data into ca 1 second size blocks (1024-point blocks for Noraxon data, 2048-point blocks for Mega-data), calculating the power spectrum and the average frequency for each block.
Noraxon
Mega
0 6 12 18 24 30 36 42 48 54 60
0 10 20 30 40 50 60 70 80 90 100
RMS EMG 100
18.48 EEkk 1000
59.136
0 kk L⋅ ⋅∆t
0 6 12 18 24 30 36 42 48 54 60
0 10 20 30 40 50 60 70 80 90 100
RMS EMG 100
11.368 EEkk 1000
57.344
0 kk L⋅ ⋅∆t
1000 S/s) we infer that we have a marked increase in the EMG-amplitude in the beginning of the Mega-data not observed in the Noraxon-data. This could be due to some difference in the execu- tions of the test?
Regr. param. Noraxon Mega
α (slope) -0.33 -0.25
β (intercept) 54.32 56.28
Average of the std.
deviation of the
frequency 2.06 1.58
In the frequency graphs we have also depicted the standard deviation of the frequenices (scaled by a factor of 10) for every block. Noraxon-data shows a little bit higher variation (see the table which also gives the averages of the deviations). Inspection of the total spectrum shows that Noraxon-data has a steeper cutoff for frequencies below 20 Hz than Mega-data. This could affect the behaviour of the calculated frequency shift. Indeed, if we have general shift of the spectrum to the lower end during fatigue and there is a high-pass for 20 Hz, then the energy will leak away for that part that is pushed below the 20 Hz limit and outside the "window" and this will reduce the observed frequency shift. Thus, the high-pass filters used by different devices may affect how they represent the fre- quency shift. Other device independent factors affecting the frequency shift are the force level in terms of MVC and the duration of the measurement. Researchers using systematically different force levels and measurement times in their work will probably get different results.
Noraxon
0 20 40 60 80 100
0.01 0.1 1 10 100100
0.01 FE
100
0 freq
Another curious observation is that if sum the non-rectified EMG-signals according to iEn)1=iEnAEMGn
N n=0,...., N B1
we get an oscillating curve with an increasing trend. Intuitively one may have exptected that this sum would have oscillated around zero supposing there are on the average an equal contributions from negative and positive potentials. The smoothed trend SE is defined by
SEn)1=0.005CSEnA0.995CiEn
Mega
0 20 40 60 80 100
0.01 0.1 1 10 100100
0.01 FE
100
0 freq
Noraxon
0 10 20 30 40
0.2 0.16 0.12 0.08 0.04 0 0.04 0.08 0.12 0.16 0.2 0.2
−0.2 iEk SEk
40
0 k L⋅ ∆⋅ t
This baselevel trend could be inherent in the electronics?
The test setup is probably good. We just need to make a greater number of trials with symmetrical conditions for Mega and Noraxon in order to find out whether there is a systematic difference. From the technical pont of view it would be also advisable to check the devices by feeding them identical artificial signal inputs and compare the characteristics of their output. I believe this has been done to some extent for a number of devices within the SENIAM-project.
Measurements: M Herrala Analysis: F Borg
Mega
0 10 20 30 40
0.2 0.16 0.12 0.08 0.04 0 0.04 0.08 0.12 0.16 0.2 0.2
−0.2 iEk SEk
40
0 k L⋅ ∆t⋅