GSR artifacts

GSR Biometrics Emotion

In this section, we will provide information about the different types of artifacts that can affect the GSR signal.

Any fluctuation in the galvanic skin response (GSR) signal that has not been caused by skin conductance changes we consider to be a GSR artifact. There are different reasons why a GSR signal can contain artifacts. For example, electrode movement will usually create rapid and large signal fluctuations, and external electrical noise will introduce a constant 50/60Hz oscillation on top of the skin conductance changes.

The best practice for obtaining reliable results is to ensure that the GSR signal does not contain any significant artifact before applying any analysis or measurement. There are two main strategies to obtain a clean GSR signal without artifacts: the first one, and always preferred, is to avoid the artifact during the recording. A good experimental setup and correct instructions to the participant can help minimize artifacts. The second one is to apply post-processing tools to detect and, when possible, eliminate the artifacts before performing an analysis or making measurements.

In the next paragraphs, we list the most common types of GSR signal artifacts and provide a few strategies on how to minimize their effects on your analysis:

High-frequency noise

The conductance of the skin changes in a time magnitude of seconds, with a frequency of 1Hz or less. Therefore, any constant fluctuation that appears in the GSR signal at a higher frequency can be considered a GSR artifact caused by high-frequency noise. The most common types are electrical noise at 50/60Hz and the precision error that the GSR sensor may introduce in the measurements. This type of GSR artifact usually has a very small amplitude compared with the GSR signal.

How to deal with it

High-frequency noise can be easily eliminated with post-processing tools by applying a noise reduction filter that removes the rapid frequency components of the GSR signal. Examples of filters that will remove high-frequency noise are a low-pass filter and a moving-mean filter.

GSR artifacts

Rapid-transient artifact

Discrete and rapid changes can also appear in a GSR signal. These rapid changes are usually much faster than skin conductance change and can have variable amplitude. A very fast electrode movement can cause this type of artifact.

How to deal with it

Rapid-transient artifacts can be eliminated with post-processing tools by applying a noise reduction filter that removes the rapid frequency components of the GSR signal. The most common type of filter is a moving-median filter.

GSR artifacts

Loose electrodes, decoupling from skin

Loose electrodes will cause electrode movement and therefore a change in the contact area of the electrode with the skin. This is likely to generate sudden rises and falls in the signal, usually larger in amplitude and faster than any rapid skin conductance response (SCR) component of the GSR signal.

How to deal with it

There are no post-processing tools that will recover the original GSR signal. The best practice is to make sure that the electrodes are correctly placed on the participant’s skin during the setup. You can use adhesive electrodes or additional tape around them to prevent the electrodes from loosening during the recording.

If you find this type of artifact in your data, the best practice is to discard that portion of data for further analysis. Otherwise, GSR algorithms can mistake the artifact and detect it as an SCR.

GSR artifacts

Electrode movement

Similarly to loosened electrodes, correctly attached electrodes can create a movement artifact if the electrode increases or decreases the contact area momentarily. Pressure on the electrodes or pulling on the electrode cables are the main reasons for this type of artifact. For example, if the electrodes are located on the participant’s hand, moving abruptly the arm or pulling on the wires will likely generate an electrode movement artifact. The signal fluctuation due to this artifact usually has amplitude and frequency components similar to a GSR signal.

How to deal with it

There are no post-processing tools that will recover the original GSR signal. The best practice is to try to avoid this type of artifact by instructing the participant to avoid abrupt movements. Placing the electrodes on the non-dominant hand and taping the electrode cables to the skin or clothes can help minimize these artifacts. If the experimental task will involve hand movements or object manipulation, it may be a good idea to place the electrodes on other body parts that are known to produce a good GSR signal, like the feet or shoulder.

Movement artifacts can have similar shapes as a GSR signal, but they can be distinguished from GSR responses by visual inspection as this type of artifact commonly has a shorter rise and fall time than a rapid SCR component. If you find this type of artifact in your data, the best practice is to discard that portion of data for further analysis. Otherwise, GSR algorithms can mistake the artifact and detect it as an SCR.

GSR artifacts

Temperature changes

Body temperature changes will impact the GSR signal, with a slow GSR signal decrease as the body temperature rises and a slow GSR signal increase as the body temperature drops. This is a physiological artifact generated by the thermoregulatory processes in our body that are not related to emotional arousal activity. If the main interest of research is the rapid changes of the GSR signal (phasic component: skin conductance responses (SCR)), this will not have a significant impact on the results. If the interest of research is focused on the GSR slow fluctuations (tonic component), temperature changes over time may mask your results.

How to deal with it

To avoid temperature changes in your participant during the experimental session, room temperature and humidity should be regulated to a comfortable level and controlled across sessions and participants.

References & Recommended Reading

(Boucsein, 2012; Roth, Dawson, & Filion, 2012)
Boucsein, W. (2012). Electrodermal activity. Springer Science & Business Media.
Roth, W. T., Dawson, M. E., & Filion, D. L. (2012). Publication recommendations for electrodermal measurements: Publication standards for EDA. Psychophysiology, 49(8), 1017–1034. https://doi.org/10.1111/j.1469-8986.2012.01384.x

Related Articles