Hi Sibylle, you don't need to change the scaling factor.  All you have to do is get an amplitude value for the maximal contraction (we use RMS for amplitude in our lab), and then get an amplitude value for the submaximal contraction.  Then simply express it as a percentage (amplitude value / MVC amplitude * 100). 

For example, lets say the maximal amplitude value was 2000 microVolts.  If the amplitude of the submaximal contraction was 500 microVolts, then you'd simply report the amplitude as being 25% MVC (don't report the raw value of 500 microVolts).

This is important because the raw EMG amplitude values vary so much across individuals.  Another subject may have a maximal amplitude of 1000 microVolts, and then an amplitude of 250 microVolts during the submaximal contraction.  Looking at the raw values, it would appear that the second subject had half of the muscle activation as the first subject (250 vs 500).  However, when normalized to MVC, both subjects were at 25% MVC, and therefore equal activation. 

To get the point of the question, all you have to do is report it as a percent to normalize it. 

Using AcqKnowledge, the ability to get a RMS amplitude value may depend on what add-on packages you have.  However, you could also just highlight the window you wish to analyze, and simply have AcqKnowledge calculate the standard deviation of the window.  If the signal is a zero mean signal (i.e. equal positive and negative deflections, with a baseline around zero), then root-mean-square (RMS) = the standard deviation of the signal.

I hope this helps.  Good luck.

I should add that the values should already be normalized to the MVC before you run your statistics.  Its not just for plotting and reporting the data.

Hi Sibylle,

I agree with Jason, but would like to add, that it could easily be done if you could export your data to excel or matlab and then do the calculation as suggested by Jason on the whole time-series.

for each point in the time-series you can calculate

(amplitude value / MVC amplitude * 100).

and then have a normalized signal.

Best,

Julian

Dear Mr. Sibylle Meier

The previous statements are accurate. For more information, if you want to analyze the difference between two or more subjects or muscles normalization techniques will be required. But, if identifying the differences between pre-test and post-test within a subject and one muscle is considering for propose of the research then it isn’t necessary to perform such an operation. However, the electrode locations shouldn’t be removing or replacing between pre-test and post-test.

Also, there are different types of MVC are utilized for the research studies like Isometric-ArbMVC and etc. choosing the accurate, more valid and reliable one is important.

Sincerely yours

Alighanbari

Mr. Mohsen Alighanbari,

I actually disagree with the statement that normalization is not necessary for within subject designs (pre vs. post).  Using the raw values will still lead to huge standard deviations, even if it is within a single group of subjects.  Using the raw values will make it substantially harder to find any statistically significant differences using standard hypothesis testing (t-tests, ANOVAs, etc.).  It could lead to an incorrect interpretation and if a real change in the data occurred, it could be missed.

My recommendation is that it is safer to just always normalize.

Sibylle was asking for scale normalization. This changes the absolute value for all data points of the raw signal.

So let's have a look at an example of two time-series from taken in the same session from the same electrode in within a short time-frame:

time-series 1: [ 1 2 3 4 2 3 5 2 4 5 1 5 5 5 1]  mV

time-series 2: [1 2 2 1 1 2 2 3 4 2 3 1 2 1 1]  mV

A paired T-Test shows a significant difference (p=0.0156) with time-series 1 being greater than time-series 2.

If the two time-series are recorded from the same electrodes with same impedances, normalization to %MVC (let's say 10 mV) for this signal would lead to a simple multiplication by the factor 10. Therefore:

normalized time-series 1: [ 10 20 30 40 20 30 50 20 40 50 10 50 50 50 10] %

normalized time-series 2: [10 20 20 10 10 20 20 30 40 20 30 10 20 10 10] %

Therefore, normalization would have absolutely no effect on within subject comparisons.

A paired T-Test again shows a significant difference (p

The way you described is interesting and classical but can linearize the EMG force relationship especially in the small force output range. Another method is to mesure the EMG RMS value obtained at the midinterval between the minimum and MVC, and to express in percent of this value. Let me know if this improve your data management

Mr. Jason M DeFreitas

Your statement is actually correct. However, what I stated before is not wrong either. Consider this situation that if the propose of our study is to identify the difference between two gait condition (with and without shoes). For acquisition of the data, electrodes were applied to the VM muscle. Then the trials were performed with this condition that the electrodes did not removed between two trials. Also, trials were performed in the single day. Therefore, for comparing the effects of two types of gait condition on VM muscle we can ignore the process of normalization technique because of flaws that this process has. But for other condition, for example difference between VM and other muscles or if the electrodes are removed, normalization technique is completely necessary and undeniable.

However, we preferred to perform normalization technique on the EMG signals in above condition too in our lab.

Sincerely yours.

Alighanbari

Thank you very much for your input and your ideas. They helped me a lot! 

Hi Sibylie

I am not sure of your application, and thus consider my response in light of other details. However, you have to be aware that the relationship between EMG features such as RMS is not linearly related with strength of muscle contraction, and RMS or similar features are very inexact measures of the muscle force. You also have to be aware of other factors that can play a very significant role such as co-contraction, which may be relevant at some levels of MVC and not others. 

Normalisation of EMG against MVC may be done for some very rough estimates, and again, the relationship would be highly subject and experimental conditions dependent, and which muscle is being considered. 

Hope that helps

If i don't have MCV in the dynamic contraction . what is the best method for that . i have the average level of the muscle activation and max peak value in this area .should i take a percentage of these data( average level / Max Peak * 100)

I got yet another question, although i posted it separately (https://www.researchgate.net/post/EMG_normalization_is_it_necessary_in_case_of_submaximal_activations)

is it necessary to normalize the sEMG signal obtained at submaximal level? I took MVC for my exercise, then asked the subject to perform exercise at 30, 45 and 60% of that MVC (maximum load subject could handle for 3 seconds). It is already normalized here. is there any need to normalize it during the evaluation of RMS?

The amplitude of EMG calculated by RMS is much smaller than the amplitude of raw EMG. This will cause that the normalized EMG will be greater than 100% when using the MVC normalization in dynamic contraction.

Hi Chuan He, I could not get your point.

in case of dynamic contractions, usually highest amplitude is taken for normalization. being MVC, the highest amplitude will definitely be more then the submaximal contractions. Also, we take the RMS of both MVC and raw signal and then normalize it.

Hi everyone, this is very interesting argoment. In my experiment i have only sEMG dinamic contraction signal (to record sEMG of lumbar spine, multifidus and lunghissimus muscles) and i have not a trial where MVC is calculated. So i took the max value recorded during the trial as normalization value [1]. This have some implication and question:

- the max value recorded during the trial can be different respect the value of RMS sEMG in MVC; how this is reflected on the test and statistical analysis?

- the relation between torque/sEMG is not linear during dinamic contraction

- i can do a comparation within subject but between subject?

- with this normalization there is lose of information for example the force of muscles? Because i normalize signal not considering max value of force that muscle can develop but using max value of sEMG produced during dinamic contraction, so i think that lose the information about the power produced by muscle

- is correct use max sEMG value of dinamic contraction to normalize the signal??

Thank you...

[1] Normalization of EMG Signals: To Normalize or

Not to Normalize and What to Normalize to? Mark Halaki and Karen Ginn

Hi Michele....

I recently came across an article (forgot the title) but it was about submaximal contractions are more suitable for normalization. But for that, one must know how much this sub-maximal is.

The key purpose of normalization is to bring all the observations to single scale. Take it like your % marks or CGPA at university. In any case, the maximum attainable value is important. we must agree upon a value beyond which our values wont go and then express the remaining values accordingly. But THAT reference maximum value should be same among all subjects (i.e. MVC or something else)

my suggestion you can used Justification amplitude and spectrum Analysia (JASA) method to measure MVC in percentage.