I am rather confused about the three options for dynamic EMG quantification process.

1. 1. FWR (full wave rectification) followed by integration; by J. Perry

2. 2. RMS (root mean sqaure)

3. 3. Linear envelope(rectification followed by low pass filter); by Vicon

Which one is better?  Or does each one of them have specific purpose?
Is something better for detecting firing moment and other thing better for comparing
amount of contraction power?
 
I hope somebody could tell me the comparative characteristics of the above three.
And related web site or title of published paper will be greatly appreciated.

From:
        "A.L.HOF" <a.l.hof@med.rug.nl>

Dear Dr Cho,
The diffference between tthe above are not worth bothering too much about.
#1 and #3 are identical. 'Integration' is in the literature often misused for 'low-pass
filtering'.
RMS, #2) is slightly different, but in practice it gives the same curves after low-pass
filtering as #1 & #3.
In fact, the choice of a sensible low-pass cutoff frequency after rectification is much more
relevant.
A good overview is in DA Winter, Biomechanics of human movement.

At Hof
Department of Medical Physiology
University of Groningen
Bloemsingel 10
NL-9712 KZ GRONINGEN
THE NETHERLANDS
Tel:   (31) 50 3632645
Fax:   (31) 50 3632751
e-mail: a.l.hof@med.rug.nl

From:
        "Plamen Gatev" <gatev@iph.bio.bas.bg>

Hi,
The main problem is not the method used but the length of  EMG epochs that is analyzed.
Shortest EMG epochs reveals muscle excitability, longer epochs- muscle power and longest epochs
muscle work.

Sincerely Yours,
Plamen Gatev MD, PhD

From:
        "Harlaar, J." <j.harlaar@azvu.nl>

        Dear Dr. Sang-hyun Cho

        basically, the linear envelope (LE) or Smoothed Rectified EMG (SRE) and the RMS of a
noisy signal with zero mean (e.g. the EMG), are representing the same property of the signal:
an estimation of the actual "intensity"of the signal. The low-pass characteristics of the filter after squaring or
rectifying the signal, determines the time period over which this "intensity"of the filter is
estimated.
        For a gaussian distribution (eg. the EMG), the RMS and the LE of a
signal x are related:

        E{|x|}=sqrt( 2/pi * E{x^2} )

        The reason that in signal-theory RMS is preferred is that it is nicely proportional to
the power of the signal (expressed in decibel: dB).  The reason that LE is used so much is that
it is easily implemented in hardware (bi-pasic rectifying).
        There also might be minor differences, though (ask a signal-processing expert).

        So use what you prefer, the low-pass filter time constant is the most critical
parameter, with regard to interpretation.

        Jaap Harlaar
        University hospital VU
        Amsterdam

From:
        David & Donna <djedjf@earthlink.net>

Dear Dr. Sang-hyun Cho

Your e-mail was forwarded to me today.  Although I have been doing EMG research for several
years now, I do not have much an answer to your question.  (It seems that the nervous system is
a rather complicated universe, with room for a great many questions!)  Anyway, Hylander and Johnson have been quite
successful in associating contractile force and EMG scored by RMS.  Their goal was to develop
statistical correlations, because EMG is easier to measure than force.

Hylander,WL; Johnson,KR (1993): Modelling relative masseter force from surface electromyograms
during mastication in non-human primates. Arch. Oral Biol. 38(3), 233-240.

Best of luck to you,

David J. Eliot
Assistant Professor
University of Bridgeport College of Chiropractic
Bridgeport, CT

From:
        Michael Dillon <m2.dillon@student.qut.edu.au>

Hello.

These two signal processing techniques perform two very different functions.

The RMS is an amplitude normalisation technique. It is one of many available and it can be
useful for comparing the relative amplitude of the signal between subjects. A good reference
for you to look at is Yang and Winter (1984) Electromyography amplitude normalisation methods:
improving their sensitivity as diagnostic tools in gait analysis. Arch Phys med rehab Vol 65,
Sept 1984.pg517-521.

A linear envelope is a signal processing technique which is used to gain an 'average trend of
EMG activity'. it is used because the RAW EMG signal is relatively useless because it
fluctuates in amplitude too quickly and too often. The linear envelop can take many forms. The
most common of which is where bins of raw EMG data are taken and averaged. This averaged figure
is then used as a representation of the activity in that bin. A bin usually is 25-50ms.   The
resultant signal is significantly smoother and hence more meaningful. This signal is usually then lowpass filtered to remove the
peaks which occur due to this processing technique to leave a smooth EMG signal. This process
of 'binning' the data and then lowpass filtering is known as a linear envlope. Winter (1990)
Biomechanics and motor control of human movement.

The linear enveloped signal in it's own right is relatively useless for determining muscle
activity. Determining periods of muscle activity is a function of many criteria. usually a
threshold is set based on some measure of muscle resting activity. The EMG signal must be able
this threshold by some arbitary measure usually 2-3 SD and some authors also specify a time
which the signal must remain above the threshold. The filter used to process the signal also
has an effect on muscle on off determination.

Hodges and Bui (1996) A comparison of computer based methods for the
determination of onset of muscle contraction using EMG. Electroencephalography and clinical
neurophysiology 101. 511-519

It would be useful for you do so some literature reviewing and look through the Biomech-l
archives as this topic is well documented even though there remains great controversy and
little consensus.

kindest regards.

Michael Dillon
B P&O Hons. PhD Student
Centre for Rehabilitation Science and Engineering
Queensland University of Technology
School of Mechanical, Manufacturing and Medical Engineering
GPO Box 2432
Brisbane. 4001.
Ph. +61 07 3864 2751 E-mail: m.dillon@qut.edu.au
Fax. +61 07 3864 1469
http://www.bee.qut.edu.au/mech/staff/mdillon.html