Thank you everyone for your responses!

I apologize for my tardy response; I wanted to perform a little piloting before I responded. We tried to solve our data analysis question again, using the responses from the post. Again we used our foam block testing dummy for these initial tests. The block was tightly secured to the vibration platform as best as possible to avoid movement. We then applied electrodes to the block in various configurations, i.e. on the side and top of the block (using surface electrodes). As we do not have an accelerometer, we used the electrodes for our testing. Each electrode was tapped and secured down to the best of our ability to the block. All cables were positioned to be clear of the movement of the plate and any other artifact. All sources of error were evaluated from the electrode all the way back to the computer output.

Once we were satisfied we had controlled for these variables as best as we could, the machine was turned on at two frequencies and amplitudes (30hz low, 30hz high, 50hz low and 50hz high). We were able to see this peak in our FFT analysis that corresponded to each frequency tested. It was also noted that the magnitude increased with the frequency tested and amplitude of vibration (displacement).

We also added weight to the plate to see if we noted a decrease in frequency due to weight alone (we hypothesized that the plate would not be able to maintain its same frequency due to this added load). We noted a decrease in at the Frequency in the FFT analysis (2-3hz decrease, example: 30hz --> 28hz). We are now in the process of repeating these steps on a subject to see how our results correspond. Further testing is needed to understand the differences between direct vibration of the muscle compared to the whole body vibration using our system and protocol.

Please find my original post as well as the responses below.

Thank you again for all of your input! You have been more than generous with your time, expertise and interest.



Abbie Ferris M.S.

Doctoral Student

University of Southern California

Musculoskeletal Biomechanics Research Laboratory (MBRL)

1540 Alcazar St, CHP-155

Los Angeles, CA 90089-9006

(323) 442-2948

Original Post:

Greetings Colleagues!

I am currently working on a pilot study involving a vibration platform. We are interested in measuring muscle activity while performing various exercises on the platform while it is turned on. Currently, we are performing an isometric squat exercise. The subject is initially standing on the platform, squats to a predetermined depth (90°), holds the squat and then the platform is turned on. We note a large increase in muscle activity immediately when the platform is turned on. However, after performing a FFT analysis, we note a single peak in the frequency spectrum coinciding with the frequency of the vibrations. The magnitude of the peak is at least 8 times that of any other frequency within the signal. Because of the magnitude of the spike at this frequency, I am inclined to consider this noise from the platform and to notch this frequency out of the signal. However, a few recent studies suggest these platforms may act to stimulate a reflex response from the muscle and thereby causing the muscle to contract at the same rate of the vibrations. I have preformed tests on a foam block strapped to the machine and placed electrodes on it and found similar results, but again, this is not a biological system and thus extrapolating the results to human subjects is difficult. Has anyone else had any experience with these platforms and EMG? Could anyone suggest a method by which we might be able to distinguish between true biological signal and noise? Any help or suggestions would be greatly appreciated and welcomed!!

Best regards!


Hello Abbie, I just ran across your message in my inbox and thought I'd send a note in case it may still be of help. Yes, tendon vibration is a common way that some folks in the neuroscience community will use to specifically target activation of muscle spindles. I use a large back massager applied to the Achilles tendons to demonstrate this effect each year in my gait class. I see two possible solutions:

1) You could collect baseline data with the subject standing quietly in an upright position with vibrationon and subtract this baseline EMG from your squat conditions. The difference should be the additional volitional muscle activity to perform the task.

2) If you can measure the exact timing of the vibrating plate and its phase relationship to the EMG signals (with no muscles), you could probably find a DSP filter that will only knock out the waveform due to this movement artifact. Any reflexive activation due to the vibration should have a time delay. This is a bit shaky and I would suggest trying #1 above first.


Young-Hui Chang

Hello Abbie,

nice protocol. You can eliminate the electronics of vibrating platform to see if your EMG has noise or not. To do this, you can take a scooter (a square platform with handles on both sides), ask your subject to squat on it and vibrate the scooter manually (with your hands) to a desired frequency. I know its a tedious job and not as accurate as vibrating platform, but I am sure it would give you atleast some idea about the noise.


Bhushan Borotikar

Graduate Student

Department of Biomedical Engineering (ND 20)

van den Bogert Lab, Lerner Research Institute

Cleveland Clinic, Cleveland OH

Dear Abbie,

I was very interested in reading your email. I am currently in just about the same position as you. I am using a Galileo 2000 vibrating platform and am conducting a study concerned with vibration effects on length-tension relationship of muscles and coactivation of agonist - antagonist muscle groups. We are concerned with the ankle plantar flexors and are looking for a possible use for vibration in rehabilitation.

We are operating at a frequency of 26Hz and upon spectral analysis of the EMG signal i find a large spike at around 26Hz as you also found.

Interestingly when i looked at the EMG spectrum of synergist and antagonist muscles to the one being vibrated there was a significantly smaller peak at 26Hz and a broader distribution of frequencies in the spectral window. I was lead to believe that the muscle should contract at the same rate of the vibrations. However i am still unsure about this matter.

I would love to hear what responses you recieved to your email and how your work is progressing.



Matthew Kemertzis

Hi Abbie,

Let me know what you find out about this. My colleague (Dr.Peter Konrad) and I have corresponded a great deal about this item and I think you have summed it up quite well.

Many who assume it's noise seem to dismiss the reflex response, yet it's precisely what we learned in year one physiology. That said, the noise likely is a factor/contributor. I'm wondering if any sort of wavelet would yield differences. One would assume the shape of the reflex response would be unique when compared to noise?

Just food for thought.

Please let me know if you find out more. :-)



Dear Abbie,

That foam block should make you think!

Very probably, your synchronous signal is the electrode artifact. With sinusoidal vibration, the electode DC potential (some 100 mV) will also be modulated sinusoidally.

If there is a reflex, this will manifest as short EMG bursts at some phase of the vibration cycle. This looks completely different.

At Hof

Dear Colleague,

I did not work with vibration platform and EMG before, but my previous work involved direct muscle vibration and EMG.

The best way to test whether the peak in EMG spectrum is noise or from reflex contraction may be that used by Lebedev and Polyakov(J Electromyography and Kinesiology, 2(1); 1992). They found that the peaks disappears during the ischemic block of the muscle, suggesting that the peaks were from reflex muscle contraction.

Another simpler way you may try is to muscle EMG when the muscle is in resting (no or very small voluntary contraction), and with the vibration being turned on for very short period (e.g. 3 second). The reason for this short duration is that reflex contraction may not have been developed.

In our previous study, we measured EMG on biceps (with no voluntary contraction) when vibration was applied to the biceps tendon for 30 seconds and we looked at the power spectrum of EMG at different stages (0 - 5s; 5-10s; 10 - 20s; 20 - 30s) after vibration was turned on. We found that there was actually no peaks in EMG at vibration frequency during 0-5 seconds after the start of vibration. However, the peak became more and more prominent in later stage (e.g. 5-10s, 10-20s and 20-30s), suggesting that more and more motor units were brought into phase-locked firing as tonic vibration reflex developed.

Hope this may help you.

Best wishes,


I think the people who will know the answer to this question are the Institute of Sound and Vibration Research, University of Southampton, UK.

Philip Helliwell,
Senior Lecturer in Rheumatology, Academic Unit of Musculoskeletal and Rehabilitation Medicine,
University of Leeds,
36, Clarendon Road,
Leeds LS2 9NZ.

I have had a problem like that before and it was do to wires crossing. I had to keep the wires from the EMG at least a foot away from the force plates to get good readings.

Hope this helps



You are almost certainly recording a tonic vibration reflex (TVR). Why should you expect it to be noise? Do you expect it to be electrical noise from the motor? In that case you also see it when the subject is standing next to the platform and not receiving the vibratory stimulus.

James Gordon, EdD, PT, FAPTA

Hi Abbie,

I assume this is surface EMG? How about if you apply the electrodes to an inanimate object - a dummy. Then if you turn on the vibrations, do you see a response in the EMG? If so, it must be noise in your EMG signal from the vibrating platform.

Also, it would be interesting to change the frequency, if you have control over that. I don't know what the top frequency that a muscle could respond to is, but at high enough frequencies, you might expect that the response of the muscle could not match the input vibrations anymore.



Dear Abbie:

I caught your posting on Biomech-L and thought I might offer some observations:

1) My first reaction to the observation is similar to yours, and I think you should pursue this issues without the use of notch filters. I am not familiar with the details of your vibrating platform, but it undoubtedly uses at least one (and possibly two) powerful motors. Electrical motors are notorious for emitting electromagnetic noise; it is inherent in their construction. Some motors are more problematic than others, depending on the type of motor, its physical construction, and its raw power. My guess is that in order to vibrate your platform, you will need to vibrate your motor at the particular frequency. Thus, the current flowing thru the motor will have this fundamental frequency component. Motors draw the most current when then start and when they stop. If the vibrational forces are established by starting and stopping the motor(s), your current, and resulting electromagnetic (EM)noise will be quite high. The EM noise can then show up in various components of your system. So the point here is that the vibrating platform needs an electrical control signal to exist at the vibrating frequency. This signal is most likely radiating and conducting past the motor to the various other component of the system and ultimately showing up in your data acquisition. It may possible to implement some straight forward noise reduction techniques on the motors and vibrating equipment itself, if this turns out to be the culprit.

2) The above argument holds even if the vibration is not generated by a motor, but by some other electromechanical device. In all cases (I think), the control signal will be electrical in nature and could find its way into your measurement system. The only case where this argument would not hold is if the vibration is generated by a pure mechanical device, like a modified bicycle. I will assume this is not the case.

3) If this frequency is a result of your equipment, then you should be able to see it under varieties of circumstances. The "debugging" phase of this question involves identified the major equipment blocks in your setup, and then isolating each one until the phenomenon is no longer observe. Lets identify some components in your system:

A) platform

B) electromechanical device vibrating the platform

C) EMG sensors and cables

D) EMG amplifier

E) Data acquisition card

F) Computer

G) the Subject

H) others??

The first test would be run all the equipment without the subject, as you I think you have done. Connect everything, including the EMG system and place the sensors on the platform. Make sure to firmly connect the EMG sensors to your EMG reference. This should give you a noise baseline reading. Acquire some data with the platform still, and then with the platform vibrating. If everything is working properly, both sets data should look very similar. You should not see 60 hz noise, and you should not see any vibration component. This is the critical test which determines what to do next.

If you see noise (which I think is the case) then you must endeavor to find the source of its "entrance". You can begin unplugging the various blocks until you no long see the noise. Be sure to set all amplifier inputs to "ref", since anything that is floating will not yield conclusive results. In this way you can determine if the noise is coming in form the EMG sensors, the EMG cables, the EMG amplifier or the A/D card itself.

If you observe no noise (with no subject) then you have a more difficult question. Is the EM noise radiation onto the subject, or is it in fact a muscular response. This is more sophisticated question which should only be addressed once the previous points have been looked at.

4)Looking at information in the frequency domain can be tricky business. I suggest you assess these signals in the time domain, since anything that is that large should be easily identifiable. A time domain analysis will yield a much more conclusive assessment of the true nature of the signal. If you can see this frequency in the time domain, it should appear to be deterministic in nature. It will be immediately obvious that its source is not physiological in nature. It will most likely be a distorted but highly consistent sine wave or square wave. The FFT is a powerful tool but must only be used under the right circumstances. For the questions that I have proposed above, a simple visual assessment should be more than adequate.

5) Critical parameters/info to understand:

A) What is your sampling frequency?

B) What is your noise baseleing (pk-pk and DC offset)?

C) What is your vibration frequency?

D) What type of A/D system are you using? (resolution, brand)

E) What type of EMG system are you using? (brand)

F) What material is the vibrating platform (conductor or insulator)?

G) Are the subjects barefoot when they step on the platform?

6) I personally am not a big fan of notch filters. I think in most circumstances, a well designed measurement system should not need them. I would suggest you try to modify your measurement system so that it is immune to external and internal noise sources. As you correctly point out, using a notch filter will also block any physiological responses that may occur at that frequency.

I would be pleased to look at any time domain data you may have if you require further assistance with this issue.

Good Luck!


Gianluca De Luca

Manager of Product Development

Delsys Inc.

617 236 0599 xt. 223

Dear Abbie

Try cross correlating the know signal from the force plate with the biological signal recorded by the electrodes. If you are getting a high correlation you can then determine that the signal is in fact coming from the vibration of the force plate.

Hoiwever, without knowing the known frequency of the force plate and the EMG signal it is hard to give advice on how to remove "noise". Just guessing you may have an aliasing error with your signal. Most EMG amps collect up to 1000Hz. If your plate vibrates at say 2000Hz the noise will be folded back over and be represented in the EMG spectrum (0-500 Hz).

The best person I can think of to speak with on this topic however woul be Jim Dickey at the University of Guelph. He is an expert on vibration/human interfaces.

Hope that helps


Cyril J. Donnelly

You may need to test the relfex on some muscles for example, arm, leg. If the same results were available all the time, you may conclude it is not noise.

During experiment, you need to take care to close the vibrations before measuing the EMG. SO, to avoid measuring the electrical signal.


Hi Abbie,

It sounds like you have movement artefact.

I had similar artefacts a while back, and interpolated across the noisy frequencies in frequency space as part of a wavelet analysis of the EMG. For more details: Wakeling, J.M., Nigg, B.M. & Rozitis, A.I. (2002). Muscle activity in the lower extremity damps the soft-tissue vibrations which occur in response to pulsed and continuous vibrations. J. Appl. Physiol. 93, 1093-1103.

Hope this helps,

James Wakeling