View Full Version : Re: Electromechanical Delay

Alan Morris
06-14-2005, 11:04 PM

Thank you Chris for starting an interesting discusssion.
I thought I would lend my two cents.

For my PhD, I have completed development of
a real-time EMG-driven musculoskeletal model of the
knee. It is driven by 12 rectified and averaged
(3 Hz FIR) EMG signals from muscles about the knee.
Knee flexion-extension repetitions of different modes
of muscle contraction (isotonic, isometric, isokinetic,
and eccentric) at diffent loads and speeds (30 -
450 deg/s) were performed and measured through
dynamometer (Biodex II) recordings.

A detailed musculoskeletal simulation model with
individual Hill type actuators was constructed.
The Hill muscle model was validated against feline
caudofemoralis data (Brown and Loeb, 1999)
where many parameters including electromechanical
delay were optimized. A value of 0.0065 second
for delay of between EMG signal and mechanical force
was found to be optimal to model the feline data.

I was able to achieve good correlations (r = 0.8
to 0.93) and low RMS error (10-16% of range)
between simulated and recorded joint moments
for isokinetic knee flexion-extension repetition trials
between 30 and 180 deg/s. The results were not as
strong for the higher isokinetic speeds: 300 and
450 deg/s, because of a more evident phase delay
between simulated and experimental data even
though the profiles were all very similar.

These results might indicate that:
1. The EMG-force relationship becomes more non-linear
at higher speeds of recruitment.
2. A fixed electromechanical delay may be accurate for
a muscle of single fibre type but the delay may be shifted
if different fibre-types are being recruited.
3. The 3Hz filter may be filtering out higher frequency
signals that are necessary to adequately model force
recruitment for higher speed activities.

More investigation with the model is necessary.


Alan Morris
Post-Doctoral Fellow
Toronto Rehabilitation Institute - Lyndhurst Centre
Bloorview MacMillan Children's Centre
Toronto, Canada

-----Original Message-----
From: * Biomechanics and Movement Science listserver on behalf of Chris Kirtley
Sent: Wed 15/06/2005 5:15 AM
Subject: Re: [BIOMCH-L] Electromechanical Delay

Just a quick reply to Ton's comments. The paper he is talking about is
"Olney SJ and Winter DA (1985) Predictions of knee and ankle moments
of force in walking from EMG and kinematic data Journal of
Biomechanics 18(1):9-20" (abstract below) but I actually wasn't really
thinking of that one in particular. There's also "White SC and Winter
(1992) Predicting muscle forces in gait from EMG signals and
musculotendon kinematics. Journal of Electromyography and Kinesiology

I haven't got Winter's book with me and these papers aren't on Science
Direct, but I think the original work he quotes was this one:
"Milner-Brown HS, Stein RB (1975)The relation between the surface
electromyogram and muscular force. J Physiol ; 246:549-569". Frank
Borg also mentions Soechting as having something to do with it.

But anyway, my point is that there is no time-shifting (such as Ton
suggests) involved - you just simply smooth the rectified EMG (i.e.
envelope it) with a bog-standard Butterworth zero-lag low-pass filter.

This is not just an academic question, because in analysing raw EMG
from patients it is assumed that the timing of electrical activity
indicates the timing of force production - I have never met any
clinician who takes into account the electromechanical delay, which is
surely present in the raw EMG. To my mind it's yet another reason why
raw EMG should not be used clinically.

Look forward to your further comments!


A deterministic model was developed and validated to calculate
instantaneous ankle and knee moments during walking using processed
EMG from representative muscles, instantaneous joint angle as a
correlate of muscle length and angular velocity as a correlate of
muscle velocity, and having available total instantaneous joint
moments for derivation of certain model parameters. A linear
regression of the moment on specifically processed EMG, recorded while
each subject performed cycled isometric calibration contractions,
yielded the constants for a basic moment-EMG relationship. Using the
resultant moment for optimization, the predicted moment was
proportionally augmented for longer muscle lengths and reduced for
shorter lengths. Similarly, the predicted moment was reduced for
shortening velocities and increased if the muscle was lengthening. The
plots of moments predicted using the full model and those calculated
from link segment mechanics followed each other quite closely. The
range of root mean square errors were: 3.2–9.5 Nm for the ankle and
4.7–13.0 Nm for the knee.

Dr. Chris Kirtley MD PhD
Associate Professor
Dept. of Biomedical Engineering
Catholic University of America
Washington DC 20064

Gait: http://www.univie.ac.at/cga
Radio: http://radiolistener.blogspot.com

Confidentiality Notice: This e-mail, including any attachments, is for the sole use of the intended recipient(s) and may contain private, confidential, and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient or this information has been inappropriately forwarded to you, please contact the sender by reply e-mail and destroy all copies of the original.