EMG signal processing for fatigue analysis by using matlab

[Eunsik Kim]

Hi everyone.


I am currently working on a piece of research for my college degree but I am finding the analysis of my data extremely difficult as I have no real knowledge of MATLAB.

The purpose of the research is fatigue analysis by using median frequency on shoulder, back, leg muscles.

I have collected raw EMG data as *txt format file.

but, I have no idea for how to handle the data to get the median frequency and how to make matlab code.

my matlab version is r2012a.

Is there anyone who can assist me??

Thank you

khemsy@naver.com

[Dan Robbins]

Re: EMG signal processing for fatigue analysis by using matlab

Hi,

MatLab can be tricky until you get your head around the basic structure of the language. If you have the signal processing toolbox you can use SPTOOL to get started and reduce the need to write code yourself. An example is covered in this PDF http://b-dig.iie.org.mx/thanassis/S/...SIP/SIP-14.pdf

An alternative would be to simply use an excel spreadsheet to calculate median frequency. This is easier to learn, though not as easy to process lots of files. you can check out Chris Kirtley's clinical gait analysis website for an example spreadsheet - this will also give you a good insight in to the process you will need to follow in MatLab if you decide to stick to MatLab.

A final point worth noting is FFT is only appropriate for analysis of isometric contractions, therefore if your protocol is not based on isometric contractions you might need to rethink your approach.

Hope that helps.

[Cory Hofmann]

Re: EMG signal processing for fatigue analysis by using matlab

Dan,

Would you be willing to provide a reference for your last statement; I would like to understand this a little more clearly. I believe I have seen papers in the past that perform different types of frequency analysis on concentric and eccentric contractions.

Many thanks,

Cory

[Dan Robbins]

Re: EMG signal processing for fatigue analysis by using matlab

Dan,

Would you be willing to provide a reference for your last statement; I would like to understand this a little more clearly. I believe I have seen papers in the past that perform different types of frequency analysis on concentric and eccentric contractions.

Many thanks,

Cory

Hi Cory,

No problem, frequency analysis during dynamic contractions has been a point of interest in our lab of late. One potential solution can be found in this article:

http://www.setantacollege.com/wp-content/uploads/Journal_db/Muscle%20Fatigue%20during%20Dynamic%20Contractions.pdf

Though whether this solves the problem seems controversial.

If you read through there is a reference to two papers indicating that FFT should be used with stationary signals. There are a few studies using FFT on dynamic contractions, however as I understand it this is not the correct approach. There are some potential applications using the STFT, or potentially wavelets. Though both of these take a little more processing time then FFT.

Hope that helps
Dan

[Matthew Tenan]

Re: EMG signal processing for fatigue analysis by using matlab

From a theoretical perspective, using the median frequency for fatigue analysis is problematic as well.

The decrease in median frequency with fatigue is supposed to be a result of decreases in conduction velocity across the sarcolemma, a point which can also be argued. The decrease in conduction velocity will cause the individual waveforms to become elongated, resulting in the FFT measuring greater power in the lower frequencies. Now, if we add in dynamic activity, which is both changing the diameter of the muscle fibers (which changes the conduction velocity of the medium) and changes the relative relationship between EMG electrode and the fibers it is measuring, you get some serious non-stationarity issues.

Most of the approaches, namely STFT, have been an effort to minimize the issues. Some people have published success using the median frequency with dynamic activity, but they're generally the exception and not the rule. You can pull some of the references from a paper I published a little while back: http://www.jelectromyographykinesiol...137-9/abstract

It is possible that the decrease in median frequency in fatigue is not due to changes in conduction velocity, which may explain conflicting findings. Either way, there is reason to be wary of the metric.

[Cory Hofmann]

Re: EMG signal processing for fatigue analysis by using matlab

Thank you both for the reply. I am more interested in the application of frequency analyses to quantify relative fast-twitch and slow-twitch muscle fiber contributions during muscle contractions. Perhaps this is an overly simplistic method with which to solve this question? Intuitively, it would make sense to me that increases in relative amplitude at higher frequencies would be more indicative of increased fast-twitch motor unit recruitment during a contraction (static or dynamic). If this were appropriate, then the fatigue induced decreases in median frequency would be a explained by 1. the increase in synchronization of slow twitch firing and 2. a reduction in fast twitch activity. I am wondering what your thoughts are on this.

Best,

Cory Hofmann

[Matthew Tenan]

Re: EMG signal processing for fatigue analysis by using matlab

People have explored this idea before (I think it is discussed in that previous link I sent because fast twitch fibers contribute more to blood lactate production).

Two thoughts on your latter points:
1. Regardless of fiber type, there is greater MU synchronization into fatigue. This has been hypothesized to be one of the contributors to tremor.
2. "A reduction in fast twitch activity" or to put it differently "fast twitch MU drop-out" has never been quantified or demonstrated (to my knowledge).

Intuitively, I understand what you are saying; however, I'm not sure surface EMG is a precise enough measure to extrapolate the metrics you are interested in. That is just my two cents, others may feel different.

[Tarkeshwar Singh]

Re: EMG signal processing for fatigue analysis by using matlab

Hi Cory,

I agree with most of the points made above. Median frequency is a good tool to measure muscle fatigue under isometric conditions but it is not the gold standard and the results should always be interpreted with caution.

The frequency content measured by surface EMG obviously depends on the neural drive to the muscles, motor unit (MU) synchronization, muscle fiber propagation velocity, MUP etc. All these factors can change during a fatiguing contraction. I would strongly recommend the three papers listed below for a detailed review. For a more comprehensive review, Gandevia 2001 is a great reference.

Lastly, I am a little apprehensive about measuring fatigue for dynamic contractions. I think wavelet-based methods might be slightly superior but I would recommend going through a lot of literature before choosing a method for analysis of dynamic contractions.

1)Vøllestad NK. Measurement of human muscle fatigue. Journal of Neuroscience Methods 74: 219-227, 1997.
2)Dimitrova NA, and Dimitrov GV. Interpretation of EMG changes with fatigue: facts, pitfalls, and fallacies. Journal of Electromyography and Kinesiology 13: 13-36, 2003.
3) Cifrek M, Medved V, Tonkovic S, and Ostojic S. Surface EMG based muscle fatigue evaluation in biomechanics. Clinical Biomechanics 24: 327-340, 2009

Tarkesh

[Cory Hofmann]

Re: EMG signal processing for fatigue analysis by using matlab

Thanks all for the references and discussion. I had the feeling that it was an overly simplistic approach.

Regards,

Cory