Biofeedback instruments supplement our awareness of biological processes ranging from the contraction of the tibialis anterior muscle in the leg to synchronous firing of cortical neurons. Hardware is designed to provide immediate and accurate information about our performance. Signals are detected, processed, quantified, and displayed. Beginners often view biofeedback equipment as intimidating black boxes that stand between them and treating their patients.

Why should a psychologist or nurse have to understand filter settings or skin-electrode impedances? Practical knowledge about these black boxes allows us to make informed purchases, use instruments effectively, and recognize malfunctions. Do you imagine that the certification examinee in the opening vignette knew when the electrodermograph worked properly or what a change in pitch signaled? If you provided biofeedback training to medical patients, wouldn’t you want to understand these details?

While this module uses the electromyograph to illustrate instrumentation concepts, most of this discussion applies directly to electroencephalographs. We will discuss instrumentation for each of the major biofeedback modalities in later modules.

This unit covers Descriptions of most commonly employed biofeedback modalities: SEMG (III-A), Sources of artifact (III-B), and Essential electronic terms and concepts for biofeedback applications (III-D).
Students completing this unit will be able to discuss:

  1. Descriptions of most commonly employed biofeedback modalities: SEMG
    A. Sensors and sensor placements
    B. Characteristic signals
    C. Signal processing and feedback displays
    D. Use of computers in biofeedback
  2. Sources of artifact
    A. How to identify and eliminate environmental noise
    B. How to evaluate instrument noise levels
    C. How to identify and correct extraneous biologic activity in recordings
    D. The relationship of skin impedance to amplifier input impedance
  3. Essential electronic terms and concepts for biofeedback applications
    A. Electrode impedance
    B. Input impedance
    C. Signal-to-noise ratio
    D. Amplifier and differential amplifier
    E. Common mode rejection
    F. Artifact
    G. Amplitude
    H. Integrator
    I. Bandpass
    J. Frequency response curve
    K. Volume conduction
    L. Time constant
    M. Integral average voltage
    N. Peak-to-peak voltage
    O. Peak voltage
    P. Root mean square voltage
    Q. Power spectrum


Biofeedback instruments measure performance directly or indirectly. A muscle sensor acts as an antenna when we directly monitor voltage sources like skeletal muscles. The main signals we directly measure include:

In contrast, a skin sensor registers temperature changes indirectly as shifts in its electrical resistance. Major signals we measure indirectly are:

Electrodes detect biological signals. They are also transducers since they convert energy from one form to another.

Consider how surface EMG (SEMG) electrodes work. Muscle fibers must become more positive inside (depolarize) before contracting. This positive shift produces a current of ions (muscle action potentials) that travels through the fluid surrounding body cells. Since interstitial fluid is a superb conductor, SEMG electrodes can detect potentials from distant motor units. This process is called volume conduction. Electrodes transform the current of ions into an electron current that flows through the cable into an electromyograph’s input jack.


Signal strength is lost during volume conduction. While the muscle action potential is measured in millivolts or thousandths of a volt, the volume-conducted signal that reaches SEMG electrodes has been reduced to microvolts or millionths of a volt, particularly by the absorption of its higher frequencies by intervening tissue (Montgomery, 2004).

The floating skin electrode is the industry standard for surface measurement of electrical signals. This design minimizes movement artifact, a false signal produced when an electrode detaches from the skin, by eliminating direct contact between metal and skin. A frontales placement is shown below.

A recessed metal disk is filled with an electrolyte, which is a conductive gel or paste. The electrolyte maintains contact between electrode metal and skin even during moderate patient movement.

We can further reduce movement artifact by securing the entire plastic electrode housing to the skin using two-sided adhesive collars and taping down the electrode cable.

Pre-gelled disposable electrodes also prevent movement artifact through their adhesive backing.

How do SEMG electrodes work? When an SEMG electrode is filled with electrolyte, the electrode metal donates ions to the electrolyte. In turn, the electrolyte contributes ions to the metal surface. Signal conduction succeeds as long as electrode and electrolyte ions are freely exchanged.

Conduction breaks down, however, when chemical reactions produce separate regions of positive and negative charge where the electrode and gel make contact. When an electrode is polarized, ion exchange is reduced, and impedance increases, weakening the signal reaching the electromyograph. This problem can result from routine clinical use. Electrode manufacturers control this problem by using silver/silver-chloride or gold electrodes which resist polarization.

At least two electrodes, designated active and reference, are needed to measure the SEMG signal. These resemble a dipole antenna used to receive FM broadcasts. We place the active electrode over a target muscle and the reference electrode over a less active site. Since our electrodes should see different amounts of SEMG activity (the active should detect more energy), a voltage should develop between them. This is the signal an electromyograph processes.

A clinician must know sufficient muscle anatomy and kinesiology to correctly position the active and reference electrodes. Active electrodes must be placed along the target muscle's belly for accurate recording. Active placement over the wrong muscle or at an angle from the muscle body will result in misleading signals (Sherman, 2002). Reference placement is less critical since it may be located within 6" of either active.

SEMG recording can be monopolar or bipolar. Monopolar recording uses one active and one reference electrode. Since the active sees more SEMG activity than the reference, a voltage develops between them.

Bipolar recording uses two actives and a shared reference electrode. Since each active is paired with the reference, bipolar recording produces two voltages. Clinicians prefer bipolar recording because it monitors a wider surface area and allows an electromyograph’s amplifier stage to remove contamination seen by both active electrodes.

SEMG scanning, where a series of muscle sites are monitored in succession, is an example of bipolar recording. Two active post electrodes share a common reference which can be the center electrode or which can be placed on a wrist strap.


Telemetry systems, like the NeXus-10 and Thought Technology Ltd.'s Tele-Infiniti telemetry systems shown below, allow clinicians, coaches, and researchers to monitor individuals during unrestricted movement. A small telemetry unit attached to an encoder's case transmits real-time information to a computer more than 30 feet away and provides immediate feedback that can be used to correct performance. The possible applications are diverse, including athletics, ergonomics, and rehabilitative medicine. For example, telemetry allows a coach to monitor a tennis player's physiological performance (SEMG, heart rate, and respiration) while delivering a high-velocity serve on an actual tennis court. This information can help athletes identify and modify maladaptive patterns of muscle recruitment and breathing that might interfere with optimal performance.



Signals like the SEMG are very weak when compared to competing false signals, or artifacts, from the body, environment, and hardware. Since the outer skin layer, or stratum corneum, consists of dead skin, oil, and dirt, it resists the propagation of the SEMG signal. We can reduce loss of SEMG signal strength by removing part of this insulating layer.

We can reduce this insulation by cleaning the skin with an alcohol wipe, gently abrading the skin with an electrode prep pad, and, if necessary, twirling electrolyte into the skin using a cue tip.


Electrode location, sensitivity, and hardware requirements will determine how extensively we prepare the skin. When finished, we should fill the recessed electrode cup until level.

Next, we need to measure the quality of skin-electrode contact which is called impedance. An impedance meter, which sends a nonpolarizing AC signal through the skin, provides the most valid measurements. In bipolar recording, we measure impedance between each active electrode and the shared reference. This results in two measurements in the Kohm (thousand-ohm) range.


An electromyograph's vulnerability to poor skin preparation depends on its input impedance and ability to filter out 60-Hz artifact. Low and balanced impedances minimize contamination by false signals. SEMG values should reflect skeletal muscle activity and not artifact from fluorescent light fixtures or power outlets.

Inadequate skin preparation and application of insufficient electroconductive gel can produce high and imbalanced impedances, and SEMG signal contamination. This problem is illustrated by the recording shown below that was generously provided by Peper, Gibney, Tylova, Harvey, and Combatalade (in press).

Caption: Recording of electrode contact artifact. When the trainee tried to relax, SEMG activity never dropped near zero due to poor skin/electrode contact. After application of electroconductive gel and the Triode™ electrode was reattached to the trainee’s non-dominant arm, the skin/electrode contact improved. The signal dropped near zero during relaxation and appropriately increased during tension.

Sherman (2002)
cautions that high impedance due to poor skin preparation can make a highly contracted muscle look virtually silent by reducing its amplitude to one-tenth of its real value.

A conservative rule is that each measurement should be less than 10 Kohms and within 5 Kohms of each other.

An impedance test can be performed manually with a separate impedance meter or voltohmmeter, or automatically by data acquisition system software.

Manual impedance testing


Automatic testing of skin-electrode contact

Gel-bridge artifact occurs when electrodes are closely spaced and the electrode gel smears, creating a bridge between the active and reference electrodes. This creates a short circuit and results in abnormally low readings. When this occurs, clinicians should remove the electrodes, clean the skin and remove the smeared gel, and then reapply the electrodes. This problem can be avoided by using pre-gelled disposable electrodes and wider electrode spacing.

The photographs shown below were generously provided byPeper, Gibney, Tylova, Harvey, and Combatalade (in press).

Caption: Narrow (Triode™) and wide (Unigel™) electrode placement.

Biological signals enter the black box via an electrode cable. Monopolar recording produces only one signal, while bipolar recording produces two different signals. SEMG signals entering an electromyograph are dropped across an input impedance, amplified, filtered, rectified, integrated, measured by a level detector, and finally displayed.


The biological signals monitored in biofeedback are very weak. The SEMG signal, for example, is measured in microvolts (millionths of a volt). These signals must first be amplified over several stages to isolate the signal we are interested in and power displays. Stereo amplifiers perform the same tasks when they boost audio signals above noise to levels that can drive loudspeakers.

The first amplification stage is called a preamplifier. A designer may place the preamplifier close to the electrodes, outside of the instrument case, to reduce signal loss as it travels down the electrode cable.

AC signals are amplified using single-ended or differential amplifiers. A single-ended amplifier is used during monopolar recording which produces only one signal. This circuit indiscriminately boosts signal, whether SEMG activity or artifact, resulting in excessive contamination.

A differential amplifier is used during bipolar recording which produces two separate signals. This design combines two single-ended amplifiers, 180o out of phase, so only the difference between the two signals is boosted. How does this reduce artifact? When there is no SEMG activity, identical noise signals reach each amplifier. The differential amplifier subtracts these signals, canceling out the artifact. The output of a perfect differential amplifier would be 0.

A differential amplifier’s separation of signal from artifact is measured by the common mode rejection ratio (CMRR). Since these amplifiers cancel out noise imperfectly, both signal and noise will be boosted. The CMRR specification compares the degree by which a differential amplifier boosts signal (differential gain) and artifact (common mode gain). CMRR = differential gain/common mode gain.

CMRR should be measured at 60-Hz where the strongest artifacts, like power line (60-Hz) noise, are found. The smallest acceptable ratio is 100 dB (100,000:1), which means that signal is boosted 100,000 times more than competing noise. State-of-the-art equipment achieves a 180 dB ratio. Lower ratios could result in unacceptable contamination of biological signals.

You can take four steps to maximize common mode rejection. First, identify artifact sources. You can use a portable electromyograph like a Geiger counter. Move the unit around the room with SEMG sensors connected, but held in your hand. Artifact sources should produce the largest display values.

Second, remove the artifact sources you find. For example, fluorescent lights can be replaced with fixtures that produce less 60-Hz noise.

Third, position the electromyograph and electrode cable to reduce artifact reception. Use the location and angle that yield the lowest readings when not attached to a patient.

Fourth, keep skin-electrode impedances balanced within 5 Kohms. If both actives receive identical noise signals, imbalance will make the signals look different and prevent complete subtraction of noise.

An amplifier’s differential input impedance further reduces the effect of unequal impedances. As SEMG signals enter the amplifier, they are dropped across a network of resistors presenting a differential input impedance in the Gohm (billion ohm) range. State-of-the-art instruments now achieve 10 Gohms. The differential input impedance must be at least 100 times skin-electrode impedance so that 99% or more of the signal can reach the electromyograph.

Why is this important? Stronger signals help an amplifier differentiate SEMG activity from noise, producing more accurate feedback.

DC signals, like temperature, are amplified using an operational amplifier (op amp). This circuit is a very high gain DC amplifier that uses external feedback to perform computations like addition, subtraction, and averaging on biological signals. Operational amplifiers are used in feedback thermometers (skin temperature) and electrodermographs (skin electrical activity).

Biofeedback instruments remove artifact through differential amplification and filters. Differential amplifiers subtract false voltages and boost signals millions of times more than noise. Filters select the frequencies we want to measure. Bandpass filters and notch filters perform this function in an electromyograph.

A bandpass filter combines two filters called high-pass and low-pass filters. A high-pass filter selects frequencies above a cutoff. In contrast, a low-pass filter selects frequencies below a cutoff. The combination of these two filters selects signals between the upper and lower cutoffs. This region is called the bandpass. For example, the 100-200 Hz bandpass filter commonly used in electromyographs combines a 100 Hz high-pass filter and a 200 Hz low-pass filter. The 100 Hz difference between the high-pass and low-pass frequencies (200 Hz-100 Hz) is this filter's bandwidth.

A bandpass filter is defined by its center frequency, cutoff frequencies, and slope. The center frequency lies in the middle of the bandpass. For a 100-200 Hz bandpass filter, the center frequency is 150 Hz. Send a 150 Hz signal into this filter and 100% will reach the next stage.

Cutoff or corner frequencies are the points where voltage is reduced to .707 of its initial strength. For an 100-200 Hz bandpass filter, the cutoffs are 100 Hz and 200 Hz. Send a 100 Hz signal into this filter and only 71% will get through.


Filter slope is the rate by which voltage is reduced as frequency changes. Slope is expressed as a ratio of decibels (logarithmic ratio of signal strength) per octave (doubling of frequencies). For example, the slopes of the high-pass and low-pass filters shown above are 20 dB/octave.

Filter slopes determine how rapidly an electromyograph excludes frequencies outside the bandpass. Steeper slopes, expressed as higher ratios, are needed when artifacts occupy frequencies near the cutoff frequencies (60-Hz noise).

Bandpass can be selected by a switch on separate electromyographs and by a switch or software on data acquisition systems. A professional can monitor narrow (100-200 Hz) or wide (20-500 Hz) frequency ranges. A wide frequency range, or bandwidth, is recommended for clinical applications since it more completely represents SEMG activity at that site. A narrow bandpass is indicated when environmental noise levels are excessive.

A wide bandwidth can better measure proportional increases in muscle contraction than a narrow bandwidth (Sherman, 2002). This is because greater muscle contraction recruits more motor units and more fast-twitch fibers, increasing the amount of signal power in the higher frequencies. While a wide bandwidth can easily detect an increase in power in the 200-500 Hz range, a narrow bandwidth can be insensitive to this change.

In the diagram below, as a client moves from muscle relaxation to strong contraction, a wide bandpass shows an increase from 2 to 9 microvolts while a narrow bandpass only shows an increase from 1 to 2 microvolts.


You can measure total noise by attaching an electrode cable to a circuit called a dummy subject. Two 10-Kohm resistors simulate the impedance of a human subject (shown below). You plug the cable with the dummy subject connected into an electromyograph which displays an estimate of noise.

The microvolt values displayed by the electromyograph represent both environmental and instrument-generated noise. The manufacturer should specify noise values for that instrument (.25 μV).

Some data acquisition systems allow clinicians to calibrate a specific SEMG sensor for a specific SEMG channel by inserting a zeroing clip. The zeroing clip shorts the sensor's inputs to allow the system to adjust for offset errors. Peper, Gibney, Tylova, Harvey, and Combatalade (in press) generously provided the photograph below.

Caption: Inserting a zeroing clip into the SEMG sensor to calibrate the equipment.

Bandpass filters reduce signal voltages outside the cutoff frequencies. Practically, these signals may still contaminate measurements to some degree. A notch filter is designed to reject a narrow range of frequencies (containing artifact) that is admitted by the bandpass filter. Sixty-Hz artifact is illustrated below.


Electromyographs are designed with nondefeatable 60-Hz notch filters that greatly reduce signals from about 58-62 Hz. This precaution preserves common mode rejection in high-noise environments, but suffers from two limitations.

First, a 60-Hz notch filter does not affect harmonics or multiples of 60-Hz (analogous to spreading ripples on a still pond) at 120 Hz, 180 Hz, and 240 Hz.

Second, desired biological signals may be eliminated along with artifact since the bulk of the SEMG signal lies between 0 and 100 Hz. While imperfect, notch filters strike a reasonable compromise between artifact reduction and signal preservation.

Skin preparation, differential amplification, bandpass selection, and notch filters collectively determine the amount of artifact contaminating the SEMG signal. The signal-to-noise ratio compares SEMG and artifact voltages. This specification should exceed 60 dB (1,000:1) at 60-Hz for adequate sensitivity, or detection of weak signals.

Sensitivity allows an electromyograph to discriminate a low SEMG value during relaxation from background noise. State-of-the-art instruments achieve ratios exceeding 100 dB (100,000:1).

The filtered SEMG signal is an AC waveform with mirror-image positive and negative halves. If we tried to measure SEMG voltage now, the sum would be 0 since the positive and negative voltages would cancel each other out.

A rectifier solves this problem by converting the filtered AC signal into a positive DC signal. Positive voltages can be added together. Manufacturers use one of two circuits. A half-wave rectifier changes the upper or lower half of an AC signal into a positive DC signal. In contrast, a full-wave rectifier converts both halves into a positive DC signal.


Rectified and filtered SEMG signals are shown below:


The rectified SEMG signal is sent to an integrator to measure signal strength in microvolts (μV). Integrators use four methods to calculate voltage:

The peak-to-peak method provides the largest estimate, equivalent to the energy contained between the positive and negative peaks of the original AC waveform, which is 2 times peak.

Peak voltage is .5 of the peak-to-peak value.

Root mean square (RMS) voltage is .707 of the peak value.

Average voltage is .637 of the peak value.

Conversion among these methods is easy. If peak-to-peak voltage is 20 μV, peak voltage is 10 μV, root mean square voltage is 7.07 μV, and average voltage is 6.37 μV.


Biomedical engineers mainly use the root mean square and average methods to measure EEG and SEMG signals. Basmajian and DeLuca have recommended the root mean square method for quantifying the SEMG. The calculated voltages are relative values and depend on both the hardware used and electrical environment.

The displayed SEMG value reflects electrode size, composition, and placement, as well as skin preparation, bandpass and notch filters, rectification method, integration method, and total noise. This means that a reading of 5 μV obtained from one electromyograph could easily be 8 μV on another. SEMG amplitude is a relative measurement, unlike temperature which is absolute in the sense that two feedback thermometers should register the same room temperature. Below is a BioGraph ® Infiniti SEMG display of microvolts RMS.

How can a clinic with several electromyograph models reduce this problem? The simplest solution is to use the same instrument with the same electrodes for a specific site and bandpass settings during all of a patient’s sessions.

An analog-to-digital (A/D) converter samples the EEG signal at a fixed sampling interval. An A/D converter's resolution is limited by the smallest signal amplitude it can sample. An A/D converter should be able to resolve amplitudes as small as 0.5 μV.

A bit number is the number of voltage levels that an A/D converter can discern. Lubar and Gunkelman (2003) consider an A/D converter resolution of 14 bits, which can discriminate among 16,384 voltage levels, acceptable. They prefer a 16-bit resolution, which can discriminate among 65,536 voltage levels. Lower A/D converter resolutions overemphasize small voltage increases.

According the Nyquist theorem, an A/D converter's sampling rate should be at least twice the highest frequency component you intend to sample. For visual inspection of the EEG, a sampling rate of 128 samples per second is acceptable, and rates of 500-1000 samples per second are preferred. Sampling at slower rates results in aliasing, where a slower beat frequency is produced by a signal near the sampling rate (Lubar & Gunkelman, 2003; Teplan, 2002; Thompson, 2003). Montgomery (2004) recommends a sampling rate that is five times the highest frequency of interest.

Mind Media's NeXus-10 claims 24-bit resolution and a sampling rate of 2048 samples per second while Thought Technology Ltd.'s ProComp Infiniti claims 14-bit resolution and a sampling rate of 2048 samples per second.

Biofeedback hardware may include a selectable time constant, which determines how long a biological signal is averaged before it is displayed. While a short time constant (0.5 s) that reveals minute, rapid changes, could be valuable in neuromuscular rehabilitation, a longer time constant (2 s) might be more appropriate for relaxation training.

In temperature biofeedback, a temperature sensor's speed is specified by its time constant. In this context, a time constant is the period required for the thermistor to reach 63.2% of a final value. You’re sitting in a 74 degree room. How long should a thermistor with a time constant of 1 s take to register a hand temperature of 92 degrees F? The thermistor will reach 99.8% of your hand temperature in 5 time constants, or 5 s. Time constants of 1 s or faster are recommended in clinical work to minimize the time lag between the temperature display and changes in blood vessel diameter.

A biofeedback therapist routinely shapes patient performance by adjusting a threshold or goal. This can be done manually or by a data acquisition system. During SEMG training, a therapist may set a starting threshold at 3 μV. The feedback tone may disappear, as a reward, when SEMG activity falls below this value. After the patient has succeeded more than 70% of the time, the therapist or software may select a harder threshold of 2.5 μV. Now, SEMG activity must fall below 2.5 μV to suspend the tone.

How does an electromyograph know when to change a display? A level detector decides whether the signal voltage matches the threshold setting to activate a feedback display.

Clinicians can now provide patients with an extensive selection of informative and motivating biofeedback displays like those from the Institute of HeartMath ®, BioTrace+ / NeXus-10, J & J Engineering, and Thought Technology Ltd.


A clinician can automatically or manually adjust the resolution (the degree of signal change) of a biofeedback screen by changing its scale (the range of values displayed). The scale should provide sufficient information to enable a client to recognize changes in signal strength and learn voluntary control. For example, in SEMG relaxation training, a 0.1 μV resolution might be superior to 0.25 μV.

A clinician can decrease the scale to keep the signal on the screen when there are large voltage fluctuations. The recordings below, which were generously provided by Peper, Gibney, Tylova, Harvey, and Combatalade (in press), show the importance of selecting the right scale.

Caption: Example of different amplification scales recorded with Triode™ electrode placed on the forearm, using a wide filter setting. Left figure shows the amplitude of the muscle tension during the tensing phase with the scale of 0 to 150 µV. However, this range would be too wide to detect any changes during the relaxation period. The right figure shows the same recording with the scale of 0 to 20 µV. In this case, you cannot see what happens to the signal during the contraction. However, you can clearly see the muscle tension during the relaxation.

Alternatively, a clinician may increase resolution when a signal shows minimal change. For example, when a client's hand temperature plateaus, a clinician may increase resolution (select x1) to better show smaller increments of temperature change. Another strategy is to manually select minimum and maximum scale values that provide a sufficiently narrow range to display signal fluctuations (75-78 o F).

Biofeedback displays include analog, logarithmic, digital, binary, and power spectral or compressed spectral array feedback.

Analog displays
vary continuously in proportion to change in signal strength. Below is a BioGraph ® Infiniti heart rate variability (HRV) display. The change in the sun's elevation provides an analog display of SDNN, which is the standard deviation of the interbeat interval. The respiration rate, amplitude, and SDNN values provide digital feedback.

In a logarithmic display, the distance between markings increases at the lower end of the scale. This approach increases resolution or ability to display a small change in signal strength at lower SEMG values. Why should designers increase resolution at the lower end? Muscle relaxation is more difficult in this range, and SEMG reductions come in smaller steps. Wider meter spacing provides more detailed feedback where it is most needed.

Designers may also provide a digital display of SEMG, like 5.1 μV. Numerical displays help clinicians track and manually record session values and may impress the patient. However, the changing digits may be more distracting to the patient than meter and light bar displays. The trend is to provide both digital and conventional analog displays on the same instrument or computer display.

A binary display shows whether the signal is above or below a threshold. Typical binary displays feature lights or tones that are either on or off. Analog displays may be more effective than binary displays since they provide the patient with more information. Which is more useful, a red light that warns you are out of gas or a gauge that constantly shows fuel level? While both displays show whether a signal has crossed the threshold, only an analog display shows the distance from the threshold.

Cutting-edge data acquisition systems now provide compressed spectral array feedback for the EEG, SEMG, and heart rate variability (HRV). Power spectral analysis measures signal amplitude (strength) across a signal’s frequency range. Fast Fourier Transform (FFT) analysis is used to separate a complex signal into component sine waves whose amplitude can be calculated. For example: a power spectral plot of the heart rate variability of a healthy subject who is breathing diaphragmatically from 5-7 breaths per minute might show a spike around 0.1 Hz in the low frequency band.

Below are EEG and SEMG displays that utilize power spectral analysis. The top BioTrace+ /NeXus-10 spectral display shows a frequency range from 1-21 Hz (X-axis) with amplitude in microvolts (Y-axis). Time progresses from the back of the display to the front. The bottom raw EEG oscilloscope display of the same sample data shows frequencies from 2-45 Hz. This movie was generously provided by John S. Anderson.

Below is a BioGraph ® Infiniti two-dimensional FFT display of frontales SEMG activity. The red bar indicates the peak frequency (highest amplitude frequency).

A biofeedback therapist routinely shapes patient performance by adjusting a threshold or goal. A therapist can manually select a threshold or allow a data acquisition system to automatically calculate one based on continuous client performance. For example, during SEMG training, a therapist may set a starting threshold at 3 μV. The feedback tone may disappear, as a reward, when SEMG activity falls below this value.

The feedback tone may disappear, as a reward, when SEMG activity falls below this value. After the patient has succeeded more than 70% of the time, the therapist or software may select a harder threshold of 2.5 μV. Now, SEMG activity must fall below 2.5 μV to suspend the tone. How does an electromyograph know when to change a display? A level detector decides whether the signal voltage matches the threshold setting to activate a feedback display.

The major artifacts that can contaminate SEMG recordings include 60-Hz, EKG, movement, and cross-talk.

Sixty-Hz artifact comes from power sources like wall outlets and fluorescent lights. The recording below from Peper, Gibney, Tylova, Harvey, and Combatalade (in press) shows the effect of turning on a light switch.

Caption: Fluorescent light artifact recorded with a Triode™ electrode placed on the left forearm extensor muscles and a bandpass filter set to 400 wide. The artifact only occurred during an abrupt change in the electrical flow. Specifically, there was only a spike in the signal when the light was turned on or off.                         

You can minimize 60-Hz artifact by equipment location, skin preparation (low and balanced impedances), bipolar recording, narrow bandpass (100-500 Hz instead of 20-500 Hz), 60-Hz notch filter, carbon-coated cables with active shielding, and selection of a well-designed electromyograph with high differential input impedance and common-mode rejection.

A BioGraph ® Infiniti display of 60-Hz artifact is shown below. Note the cyclical fluctuation in SEMG voltage.

EKG artifact results when the R-wave is detected by sensors placed on the upper limbs or trunk (a wide trapezius placement is particularly vulnerable). Peper, Gibney, Tylova, Harvey, and Combatalade (in press)generously provided the photograph shown below.

Caption: Wide upper trapezius electrode placement with two active electrodes placed on the center of the right and left upper trapezius muscles and the reference electrode placed on T1 of the spine.

The frequency range for this artifact is .05-80 Hz. EKG contamination is seen in rhythmic meter fluctuations in the wide-bandwidth signal (20-500 Hz) displayed below. This demonstrates why clinicians must visually inspect the analog signal instead of only examining digital values. The recording below from Peper, Gibney, Tylova, Harvey, and Combatalade (in press) shows that a narrow bandwidth (100-200 Hz) can minimize EKG artifact.

Caption: Effects of filter setting on EKG artifact recorded from the upper left trapezius Triode™ SEMG.

Below is a BioGraph ® Infiniti display of EKG artifact.

You can minimize EKG artifact by locating sensors diagonally from the heart, spaced 1 cm apart, and using an EKG notch filter and a lower frequency cutoff around 100 Hz. Simply narrowing a 20-500 Hz bandpass to 100-200 Hz, without relocating the SEMG sensors, eliminated EKG artifact from the BioGraph ® Infiniti display below.

Radio frequency artifact
radiates outward like a cone from the front of televisions and computer monitors. The recording below from Peper, Gibney, Tylova, Harvey, and Combatalade (in press) shows the effect of a cell phone on the SEMG signal.

Caption: SEMG recording with Triode™ electrodes from the forearm extensors. There is a significant artifact from placing a cell phone near the Myoscan sensor and turning it on. The artifact disappears if the cell phone is further than 30 cm from the sensor.

You can control radio frequency artifact using the following steps: (1) move equipment away from high-frequency artifact sources (radiology equipment), placing biofeedback equipment behind or to the side of monitor and never closer than two feet in front (Montgomery, 2004); (2) use a 40-Hz high-pass filter, differential amplifier, differential input impedance over 1 megohm, and fiber optic electrode cables; (3) maintain low and balanced skin-electrode impedance; and (4) install a Faraday cage.

Electrostatic artifact is produced by static electricity, often encountered in low-humidity environments. This artifact can be minimized by using electrostatic floor mats and sprays which can be applied to vulnerable peripherals like keyboards. A good precaution is to discharge static on a grounded object containing metal (cabinet or water pipe) before touching biofeedback equipment.

Movement artifact occurs when patient movement displaces the electrode cable. You should suspect this artifact when you see unexpectedly elevated values or high-amplitude waveforms. Note the large-scale voltage changes after 192.0 and 195.0 seconds due to electrode cable movement in the blue tracing shown below on this BioGraph ® Infiniti display.

Movement artifact can be controlled by instructing your patient to limit movement, observing your patient to ensure compliance, and securing the sensor with adhesive collars and the cable to a chair with tape at several points. 


Cross-talk artifact occurs when adjacent SEMG activity contaminates your readings. You can detect cross-talk by monitoring an adjacent muscle with a second SEMG channel. Cross-talk artifact can be reduced by locating electrodes along muscle striations, closely spacing the electrodes, and positioning the patient’s body so that antagonist muscles are inactive.


Data acquisition software has become increasingly sophisticated and can now assist clinicians in artifact detection and removal. For example, BioGraph ® Infiniti software automatically detects artifacts and highlights them for visual inspection and removal. After clinicians have removed contaminated data segments, they can calculate accurate session statistics from the remaining data.

You can determine whether an SEMG display mirrors your client's muscle contraction by performing a tracking test, during which you instruct your client to briefly contract and then relax the monitored muscles. For example, for a frontal placement over the forehead, you might ask your client: "Please gently tighten the muscles in your forehead for a few seconds and then allow them to relax." The integrated SEMG signal should increase during the contraction phase and decrease during the relaxation phase.

Below is a BioGraph ® Infiniti SEMG tracking test display in which the client briefly contracts and relaxes the frontales muscles three times in succession. See the SEMG signal peak three times.

A tracking test checks the integrity of the entire signal chain from the three individual sensors to the encoder, and the correct software selection of input channels.

The integration of computers into biofeedback has allowed more powerful data analysis, like Fast-Fourier Transformation to provide power spectral analysis of complex EEG, SEMG, and heart rate variability (HRV) signals, more powerful displays to better train patients, and turnkey record keeping. Computers also enable clinicians to administer and score psychological tests like the MMPI-2 and TOVA.

A BioGraph ® Infiniti SEMG 3-D Fast Fourier Transformation display is shown below. 

The downside of computer use is that they are sources of 60 Hz artifact and radio frequency artifact, and can present a shock hazard without proper electrical isolation.

Now that you have completed this module, find the manual for your electromyograph, electroencephalograph, or data acquisition system. Find the following information:
Common mode rejection ratio:
Input impedance:
Integration method:
Signal-to-noise ratio:

Perform the behavioral test discussed in this module. Place a set of SEMG electrodes on the palmar surface (underside) of your arm and then watch the display as you contract and relax these muscles.

If you have a portable electromyograph, insert an electrode and move it about the room to detect the lowest and highest zones for 60-Hz noise.

Amar, P. B., McKee, M. G., Peavey, B. S., Schneider, C. J., Sherman, R. A., & Sterman, M. B., (1992). Standards and guidelines for biofeedback applications in psychophysiological self-regulation. Boulder, CO: Association for Applied Psychophysiology and Biofeedback.

Andreassi, J. L. (2000). Psychophysiology: Human behavior and physiological response. Hillsdale, NJ: Lawrence Erlbaum and Associates, Inc.

Basmajian, J. V. (Ed.). (1989). Biofeedback: Principles and practice for clinicians. New York: Williams & Wilkins.

Cacioppo, J. T., & Tassinary, L. G. (Eds.). (1990). Principles of psychophysiology. New York: Cambridge University Press.

Carr, J. J., & Brown, J. M. (1981). Introduction to biomedical equipment technology. New York: John Wiley & Sons.

Cromwell, L., Weibell, F. J., & Pfeiffer, E. A. (1980). Biomedical instrumentation and measurements. Englewood Cliffs, NJ: Prentice-Hall, Inc.

Lubar, J., & Gunkelman, J. (2003). Neurometrics, neurotherapy, and clinical practice. Workshop presented at the 34th annual Association for Applied Psychophysiology and Biofeedback convention, Jacksonville, Florida.

Montgomery, D. (2004). Introduction to biofeedback. Module 3: Psychophysiological recording. Wheat Ridge, CO: AAPB.

Peavey, B. (2003). Effects of drugs on biofeedback. Short course presented at the 34th annual Association for Applied Psychophysiology and Biofeedback convention, Jacksonville, Florida.

M. S. Schwartz, & F. Andrasik (Eds.). (2003). Biofeedback: A practitioner's guide (3rd ed.). New York: The Guilford Press.

Peper, E., Gibney, K. H., Tylova, H., Harvey, R., & Combatalade, D. (in press). Mastery through experience..

Sherman, R. (2002). Hooray! The revolution is here! (But don't stop it in its tracks). Biofeedback, 30(1), 7, 18.

Stern, R. M., Ray, W. J., & Quigley, K. S. (2001). Psychophysiological recording (2nd ed.). New York: Oxford University Press.

Teplan, M. (2002). Fundamentals of EEG measurement. Measurement Science Review, 2(2), 1-11.