GSR data filtering and analysis in Pro Lab

GSR Tobii Pro Lab Shimmer

Identifying skin conductance responses (SCRs) and extracting their main characteristics is a common practice in galvanic skin response (GSR) research. SCRs can be produced in response to a specific event (e.g., stimulus onset), known as event-related skin conductance responses (ER-SCR), or appear spontaneously with varying rates. This page will both explain Pro Lab’s procedure for identifying SCRs and ER-SCRs in a GSR signal, and describe the main SCR and ER-SCR characteristics that Pro Lab calculates.

Pro Lab GSR analysis is based on a three-step approach:

1. GSR data filtering:

The GSR signal varies slowly in time and therefore, any rapid change in the GSR signal is considered external noise. In order to analyze the GSR data, it is important to remove first the most common types of noise or artifacts: high-frequency noise and rapid-transient artifacts. Pro Lab will remove these types of artifacts by applying a median filter with a time window of 500ms, followed by a mean filter with a time window of 1000ms.

GSR analysis

2. SCR detection analysis

After data filtering, Pro Lab applies a SCR detection algorithm to identify SCRs in the GSR data and calculate their main characteristics. Pro Lab follows these steps:

  1. Pro Lab downsamples the GSR by an integer factor (only samples are deleted, no interpolation). In the case of Shimmer3 GSR+, the downsample factor is 8, having a final sampling rate of approximately 15Hz.
  2. Pro Lab applies a standard peak detection method (trough-to-peak) to identify all local maxima and minima of the GSR signal. All trough-to-peak pairs are classified as SCRs if their amplitude is higher than the minimum amplitude threshold of 0.03µS.
  3. For each SCR detected, Pro Lab calculates the main characteristics of an SCR:
  • SCR amplitude: Amplitude difference between GSR level at SCR onset and GSR level at SCR peak.
  • SCR rise time: Time difference between SCR peak time and SCR onset time.
  • SCR half recovery time: Time difference between the time when the GSR level has recovered to 50% of the SCR amplitude and the SCR peak time. In two cases Pro Lab will not calculate this characteristic for a certain SCR (SCR-1): (1) when a second SCR (SCR-2) starts before the GSR level has recovered to 50% of the SCR-1, or (2) if the recording finishes before the GSR level has recovered to 50% of SCR-1.

3. ER-SCR classification

An SCR can be considered to be produced in response to a specific event (e.g., stimulus onset) when the SCR onset falls within a specific time window after the event (Boucsein, 2012). In Pro Lab, all identified SCRs with their SCR onset falling within 1s to 5s after any stimulus onset and custom Time of interest (TOI) interval start will be classified as ER-SCRs.

For each ER-SCR, Pro Lab will calculate:

  • Event: Name of the event that caused the ER-SCR.
  • Latency: Time difference between the SCR onset time and the event time.

Additional notes:

  • There will always be a maximum of one ER-SCR for each event. If more than one SCR onset is found within the time window, the first SCR will be considered the ER-SCR.
  • If no SCR onset is found within the time window of an event, the event will be considered as not having elicited an ER-SCR, and will be indicated in the GSR metrics.
  • An SCR can be classified as an ER-SCR for more than one event. One example of this is when a stimulus is shown for less than 4s. In this case, there will be two stimuli onset in periods shorter than the ER-SCR time window. Pro Lab will classify the SCR as ER-SCR for the two events. In the replay of individual recordings, this will be displayed as more than one blue marker on top of the ER-SCR peak.
GSR analysis

References and Recommended Reading

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