EMG Parameters — Which Number Answers Which Question
A surface EMG recording can answer five kinds of question — is the muscle active at all, is it more or less active than in another condition, when does it switch on and off relative to others, how much of its capacity does it use, and does it fatigue (Konrad's five question levels). A sixth question, how the muscle coordinates with its neighbours, cuts across all five (Konrad p. 51: coordination can be examined on every level, with at least two muscles). Each question has its own parameter: raw-curve inspection, mean amplitude, onset/offset thresholds with time-to-peak and firing order, %MVC, median-frequency slope, co-activation indices. Choosing the parameter before the recording — not after — is what makes an EMG measurement interpretable.
Where exactly does the electrode go?
See Vastus Medialis on the free 3D model — electrode position, orientation and how to find it.
Open in 3D
Why start with the question
Konrad's analysis chapter is built around a simple observation: the parameters are cheap, the question is expensive. Software will compute mean, peak, area, median frequency and onsets for every channel in a second — but a mean amplitude cannot answer a timing question, a peak cannot answer a fatigue question, and a single-channel value cannot answer a coordination question. Konrad (p. 45) makes the same point from the practical side: the higher the question level, the more complex the analysis — a one-channel raw recording with a minute of preparation answers "is the muscle active?", while the coordination quality of complex movements may call for 16 channels, MVC normalisation and an hour of preparation. So decide what you want to know first, then choose the placement, the task and the number.
Every parameter below assumes the processing chain from raw EMG to RMS and, where amplitudes are compared across sessions or people, normalisation.
Level 1 — Active or not?
Question: Is the muscle active at all during this task?
Parameter: a yes/no answer from the raw curve — inspect the raw signal against a clean baseline. Konrad (p. 45): this needs only one channel and about a minute of preparation. A numerical threshold is not required until you ask when (Level 3).
Rule: check the baseline before you judge — on a noisy or drifting baseline, "active" has no meaning.
Level 2 — More or less?
Question: Is the muscle more active in condition A than in B — left vs. right, before vs. after, exercise 1 vs. exercise 2?
Parameter: the mean amplitude of the envelope over a defined interval — Konrad's "probably the most important EMG calculation": robust, least sensitive to interval length, best for comparisons. Peak is too variable on single trials; use the average peak (mean of the ten highest values) or peaks of averaged curves. Area (iEMG) only at equal interval lengths.
Rules: same electrodes, same task, same window; for left–right comparisons within one session raw µV are acceptable if placement and preparation are identical — allow roughly 10–15 % inherent variance from differing detection conditions (Konrad); across sessions or people normalise. And remember Cram's warning: more microvolts is not more force.
Level 3 — When? (timing)
Question: When does the muscle switch on and off, in which order do the muscles fire, how long after a stimulus, when is the peak?
Parameters: onset and offset, defined by a threshold on the envelope, plus time to peak (from contraction start or event to peak of the envelope), firing order across channels (onset ranking), onset pattern diagrams (on/off bars per muscle across a movement cycle — the classic gait representation), and reaction time after an external trigger. Konrad lists three ways to set the onset threshold:
- Multiple of baseline SD — typically 2–3 standard deviations of the resting baseline, plus a minimum duration (e.g. 50 ms) above threshold so single spikes do not trigger. Popular, but the baseline SD varies between trials and people; with very quiet amplifiers the factor may need to rise to 8 or more.
- Percentage of local peak — e.g. 5 % of the peak within the analysis window; more stable, independent of baseline noise.
- Fixed value — a microvolt level, or better a % MVC level in normalised data.
Whatever the rule: check every onset graphically. Hodges & Bui compared 27 onset methods differing in filter, threshold and duration criterion; most deviated significantly from the visually determined onset, so the detected onset moves with the method — a timing conclusion is only as good as its threshold.
Rules: short smoothing windows or zero-phase filters (every moving window delays onsets), a synchronised event marker (foot switch, goniometer, video), and identical threshold rules for all channels. Typical clinical questions: does the Vastus Medialis start before the Vastus Lateralis? Does the Gluteus Maximus fire before the Biceps Femoris (Long Head) in hip extension? Does the Multifidus switch on in a feed-forward manner before the arm moves?
Level 4 — How much?
Question: How much of its capacity is the muscle using?
Parameter: mean amplitude expressed as % MVC (or another reference — see the normalisation article). Only normalised data can say "the Upper Trapezius worked at 15 % of maximum during this typing task". Konrad's Input % — each channel's mean as a share of the summed means — answers the related question "how is the effort distributed across these muscles?"
Rules: the reference contraction must be valid (see MVC test rules); values above 100 % occur in dynamic tasks and are a reason to check the reference, not to clip.
Level 5 — Does it fatigue?
Question: Does the muscle show signs of local fatigue during a sustained effort?
Parameters: in a static, submaximal contraction at constant load and fixed joint angle, two things happen over time (Konrad, p. 50): the amplitude rises (additional motor units are recruited) and the mean and median frequency of the power spectrum fall (conduction velocity slows; the spectrum shifts left). A falling median frequency over time is the classic spectral marker of muscle fatigue (overview: De Luca 1997).
Rules: the spectral method needs stationarity — a constant, static contraction. In dynamic tasks (strength training) use the amplitude rise instead. Watch for the opposite pattern (frequency up, amplitude down), which Konrad attributes to load shifting to synergists or reduced antagonist co-activation. Applications: back-endurance tests (multifidus and erector spinae in low-back pain), evaluating training effects.
Across all levels — Coordination
Question: Do the muscles around a joint or along a chain work together appropriately? Konrad (p. 51): this can be investigated on all five levels, but needs at least two muscles.
Parameters: all of the above, but on at least two channels and with explicit criteria. Konrad's examples of "good" coordination: symmetrical innervation of synergists (the vasti), synchronised firing order within a chain, feed-forward activation of stabilisers, and antagonist co-innervation that is low and late. Co-activation is quantified from the overlap of agonist and antagonist envelopes; ratios such as VMO/VL or upper/lower trapezius are the clinical shorthand.
Rules: define the criterion before recording; record synergists and antagonists together; watch for crosstalk masquerading as co-activation (a channel that mirrors its neighbour perfectly is suspicious).
The map
| Question | Parameter | Needs | Typical trap |
|---|---|---|---|
| Active? | yes/no from the raw curve | one channel, clean baseline | noisy baseline read as activity |
| More/less? | mean amplitude of envelope | same setup; normalise across sessions/people | reading µV as force |
| When? | onset/offset by threshold, time to peak, firing order, onset pattern | defined threshold, visual check, event marker, zero-phase smoothing | threshold rule or window lag shifts onsets |
| How much? | % MVC (mean) | valid reference contraction | poor MVC → wrong % |
| Fatigue? | median-frequency slope; amplitude rise | static, constant load; stationarity | dynamic task → spectrum meaningless |
| Coordination? (all levels) | ratios, co-activation, firing order | ≥ 2 channels, explicit criteria | crosstalk read as co-activation |
The five levels are Konrad's question levels; the "six parameters" in the figure below are measured quantities of a single burst, not another set of levels.
Muscles referenced in this article
From placement to measurement
EMG Guide shows where the electrodes go. easyEMG with PicoBlue sensors visualises the signal live afterwards — SENIAM-conform, in real time, with a signal check on screen.
Frequently asked questions
Which single parameter should I use if I can only report one?
The mean amplitude of the RMS envelope over a defined interval, normalised if you compare across sessions or people. It is the most robust number and answers the most common question ("more or less").
Why not use the peak?
Because on a single trial the peak is dominated by the random superposition of motor unit potentials — it is not reproducible. Use the average peak or the peak of an averaged curve.
Can I measure fatigue during a dynamic exercise?
The spectral fatigue index needs a static, constant contraction. In dynamic tasks the amplitude increase at constant external load is the usable indicator, with the caveats Konrad gives.
How many channels do I need for a coordination question?
At least the agonist and one antagonist or synergist; for a chain (e.g. hip extension), all main contributors — the Gluteus Maximus, hamstrings and lumbar extensors. One channel cannot answer a coordination question by definition.
Sources
- Konrad P. The ABC of EMG. Noraxon; 2005 — "EMG Analysis": Analysis Questions Overview (p. 45), On/Off (p. 46), More/Less (p. 47), Muscle Timing (p. 48), How Much Activity (p. 49), How Much Fatigue (p. 50), Movement Coordination (p. 51); Amplitude/Frequency/Timing Parameters (pp. 39–43).
- De Luca CJ. The use of surface electromyography in biomechanics. J Appl Biomech. 1997;13:135–163.
- Hodges PW, Bui BH. A comparison of computer-based methods for the determination of onset of muscle contraction using electromyography. Electroencephalogr Clin Neurophysiol. 1996;101:511–519.
- Criswell E. Cram's Introduction to Surface Electromyography. 2nd ed. 2011.
- Merletti R, Parker PA (eds.). Electromyography. IEEE/Wiley; 2004.
Read next
10 Factors That Distort Your EMG Signal
Skin, fat, temperature, spacing, innervation zone, crosstalk, ECG, motion, cables, fatigue — ten influences on EMG amplitude and how to control each.
From Raw EMG to RMS — Rectification, Smoothing and Filtering Explained
What happens between the raw EMG trace and the amplitude number: rectification, moving average vs. RMS, window length, notch filters.
MVC Normalisation — Why Microvolts Alone Say Nothing
How to normalise surface EMG to a maximum voluntary contraction (MVC/MVIC), how to run the reference test, when MVC fails, and the alternatives.
All articles →