Hello, I want to use Noraxon equipment to measure the EMG signal of masticatory muscles in living humans performing isometric biting tasks to a submaximal bite force, which last on average 1.5 seconds. I have read several articles to define the protocol and parameters to filter and process the signal. but they vary a lot. My questions are:
- When an initial filter has been set during the recording phase (let's say 500 Hz), how do I know if a new filter is necessary at the smoothing phase?
- In preliminary tests, the time before and after peak activity (as calculated by the system) was 100 ms each. So, if I understood right from literature I think I should use 200 ms to measure muscle activity. Is this correct?
- If I choose 200 ms, and given that each muscle contraction lasts for 1.5 s, How should I pick the part of the burst where to take those 200 ms from? could I just use a longer period (e.g 1 s)?
- If I smooth the signal using rms over a long period of time (500 ms), I get a nice curve which would allow me to pick the peak of muscle activity for my analysis. St-Amant et al (Proceedings of the 1996 IEEE Twenty-Second Annual Northeast, p. 93-94) say that this long period gives a better signal-to-noise ratio. Is this ok in my case? how can I estimate how much time to use for rms smoothing?
- Which role plays the "sampling frequency" in an analysis as the one I have presented here? Some articles talk about it, but it is not general, and I don't really understand if it is just an alternative method to define periods or it is indeed a mandatory step in the analysis?
Dear Viviana
First of all, I am interested in knowing your recording protocol precisely. I also worked on sEMG signal recording from masticatory muscles (Temporalis) during rest and clenching. But I had serious problems during recording:
1. Subjects with denture! or filled or extracted molar teeth. It may seem funny, but affects my results a lot.
2. Psychological states of subjects. Stress, anxiety, special disease such as headache may have an important role on the amount of pressure that one will produce either in rest state or MVC. You just mentioned living humans, are they healthy? (mentally or physically)
Because of such problems, it seems that subjects should be chosen carefully, moreover not forget the normalization. Normalization of MVCs with respect to the rest signals from each subject, may help you to reduce the big variations.
The second point is that as the biosignals are continuous, so the recording equipments should make them discrete. Thus at this time sampling frequency would have a Decisive role to maintain as much information as possible due to discretization. Nyquist rate says that it should be twice the maximum frequency existed in the signal, namely 2*500 Hz. But due to the lack of an ideal (anti aliasing) filter, sampling frequency is chosen between 3 and 11 times of max frequency. If I were you, I would choose 2KHz (2000 Hz).
Viviana, unfortunately I could not get what you mean by questions 2,3, and 4. Is it possible to explain more about your methodolody and of course the aim and motivation of your study? (What is the purpose of sEMG recording?)
And about additional filtering, I guess the filters that you use are hardware filters of Noraxon, right?
After recording you can calculate FFT of your signal to see the performance of your previous filters and also to know if additional filtering is necessary.
Hope this information can be useful to you.
Fariba
Dear Fariba,
Your reply gives me more clues about the next steps. Still I am not clear about some things. My specific aim is to relate muscle activity with bite force for two specific, healthy individuals (not for group comparisons).
Aquisition was made with a Noraxon equipment, with an initial filter of 500Hz, a gain of 1X and a sampling rate of 1000 Hz. Surface electrodes were placed in standard locations to measure anterior temporalis and masseter activities. Tthe individuals performed bites of 1.5 s with pauses of 5 s , and at the same time the bite force was recorded.
My problem has been mostly about how to process the signal:
1- In several articles, I have read that addtional filters are used. How do I know when this is necessary?
2- About the signal smoothing by RMS, how do I choose the time window?, and related to this, how do I choose the time window to actually calculate muscle activity?. Do these two decisions relate to each other? (eg, does the smoothing window need to be X times lower or higher than the window foir activity measurement?)
3- About sampling rate, can this be modified after data collection or every time I use a different sampling rate I have to measure new data? I ask this because it seems that additional filters can be used after the signal is recorded, but I am not sure if this works also for sampling rate.
I hope I am clear enough, and thanks for the reply.
Best wishes,
Viviana
Hello,
1- Actually filters are used for different purposes. Some times we use them for feature extraction for example low pass filters (LPF) for extracting the signal envelope. So after signal recording, you should perform additional LP filtering. On the other hand, there are some situations in which, after recording the signal in a specific environment and with a system such as Noraxon, you see that the signal quality in terms of signal to noise ratio or specific noises such as power line noise, is not acceptable. How can we understand it?
Well, we can import the recorded signals to another software environment such as MATLAB, or even in Noraxon environment, then calculate the power spectrum of the signal. If, for example there are several frequency components above 500 Hz, it is necessary to use additional LPF.
2- As we know, most biosignals are highly nonstationary even for isometric tasks. So, we can not consider the whole signal and perform the analyses. First of all, it is suggested to do signal framing. The frame length should be so that we can consider the signal to be stationary in it (namely, its statistical properties do not change with time in that frame). Frame lengths such as 250 ms, 500 ms are ok. This frame length can be used for both smoothing and feature extraction, in my view.
In addition, because your task is not fatiguing (1.5 second) and the state of your signal does not change, I think you should not worry about the exact length of windows. By the way, about this question if I had the article that you read, then we could discuss about it better. Could you please attach it for me? Thanks in advance.
3- You are right. As mentioned before, additional filters can be used after signal recording. Also it is possible to change sampling frequency after signal recording. For example comments like " downsample" or "upsample" can be used in MATLAB software. I used PowerLab system for my recordings. This system does not have facilities to change the sampling frequency after data collection, so I had to import the signals to MATLAB and then perform up/down sampling. But, I think it is not good to change the sampling frequency (specially upsampling) that you used at the time of recording. Please see the attached file. This file shows 5 seconds of EMG signal recorded from Temporalis muscle during rest. The left one is the signal with sampling frequency of 1000 Hz and the right one is the same signal but I changed the sampling frequency in MATLAB software after recording. You can see the adverse effects of upsampling clearly.
Again, I hope this ideas be informative.
All the best,
Fariba
Hi Fariba,
Thanks a lot, I think I have more clarity now and ideas to proceed. Please find attached the article I mentioned. Hope it is useful for you.
Best wishes,
Viviana
- Badges
- Science topic
- Similar topics
- Cell Biology
- Methods