Guide to Metric Calculations

Understanding the decoding of raw data into metrics

All metrics are calculated in a personalized way, using each person’s initial physiological state as a reference.

Regardless of the technology used, there is no standardization that allows for a uniform measurement of consumers' emotional responses. The calibration is necessary to calculate our metrics.

Calibration is key to the calculation of any metric: it transforms raw data into indicators tailored to each individual and enables direct comparisons between people.
Thanks to this process, we can measure the participant’s physiological state at a specific time and place, providing more accurate and contextualized results. Each result reflects changes relative to that person’s natural state, not absolute values. Metrics scores are always expressed in relation to the calibration value.

In this post you cand find and calculation of all metrics related to the hardware used:

Ring GSR/BVP metrics

Activation:

Raw EDA signal capture: The sensor (Ring) continuously records the skin’s electrical resistance. This depends on the moisture produced by sweat glands, which are controlled by the autonomic nervous system.
Low-pass filtering: Fast fluctuations are removed to retain only slow and sustained changes typical of the Skin Conductance Level (SCL). Activation is measured over long periods (10 seconds or more).
Removal of motion artifacts: The accelerometer detects hand movements, and the algorithm excludes parts of the signal that do not correspond to emotional responses.
Referencing to task start: The initial SCL value (baseline before the task begins) is subtracted from the signal. This allows focusing on relative changes during the task rather than absolute values, which vary between individuals.

Impact:

Component separation: The EDA signal is decomposed into a slow component (SCL) for activation and a fast component (SCR), which includes sudden conductance peaks for measuring impact.
Response identification: All individual peaks caused by rapid body reactions to stimuli are detected. This can include startle responses, sudden interest, surprise, and more.
SCR average calculation by task: The average intensity of the detected peaks is calculated for each task.

*Impact is calculated as a binary response in the Passive Image Template.  This information is meant to be useful in the aggregate (averaged over multiple participants). 

EEG signal processing

How is the signal processed:

EEG signal cleaning: Before analyzing emotional responses, it is essential to ensure the signal is clean. The following process is applied to all metrics:
Filtering of relevant frequencies: Frequencies that do not carry emotional information (extremely low or high) are removed.
Removal of strong noise: Moments with significant noise (e.g., due to movement or interference) are detected and corrected by reconstructing the signal using information from nearby channels.
Source separation: The signal is decomposed into different components to identify which parts come from brain activity and which are noise (such as blinks or muscle tension).
Personalization of brainwave: Each brain is unique, so the analysis bands are adjusted individually for each person.
Identification of the “Alpha Peak”: The activity with eyes closed and open is compared to identify the peak frequency.
Definition of personalized bands: Based on the alpha peak, three personalized bands are defined: Alpha, Theta, and Beta.
Personalized normalization: For each brainwave band, all signals related to the subject’s specific metric are collected, and the 25th and 75th percentiles are calculated.

Diadem EEG metrics:

Valence:

Measurement of frontal alpha activity: The amount of electrical activity in the alpha band is measured in two areas of the brain (F3 on the left and F4 on the right).

*Valence is calculated somewhat differently in the Passive Image Template and requires some self-reporting for accurate calculation. 

Memorization:

Theta band focus
Identification of the personalized theta band: Activity is analyzed with eyes closed and open to find the alpha frequency, which is then used to define the personalized theta band (related to memory).
Filtering the signal in the theta band: Only the part of the signal corresponding to the personalized theta band is extracted.
Measurement of activity in frontal areas: The amount of energy in the theta band is measured in specific brain areas (Fp1, Fp2, F7, F3, F4, and F8).

Workload:

Identification of the “Alpha Peak”: Brain activity with eyes closed and open is analyzed to determine the frequency at which the brain is most relaxed.
Mental effort measurement: The energy in brainwaves associated with mental effort is calculated:
- Frontal Theta (areas F3 and F4)
- Parietal Alpha (areas P3 and P4)

Engagement:

Measurement of mental engagement: Brain energy is measured in areas F3, F4, P3, and P4 across three frequency bands:
- Beta: Associated with concentration and alertness, depending on the context.
- Alpha and Theta: Associated with relaxation and disengagement, depending on the context.

Tobii Eye-tracker

Eye tracking:

Tobii Eye-tracking metrics in Sennsmetrics are only provided when an Area of Interest (AOI) is defined on the image. All metrics are calculated based on fixations.

Fixation: The fixation calculates where the person is fixing the gaze. Can be defined as the periods of time where the eyes are relatively still, holding the central foveal vision in place so that the visual system can take in detailed information about what is being looked at. Fixation durations typically range from 50 to 600 milliseconds, but this can vary depending on the task and visual information. To calculate it, the Tobii algorithm is used.

For more information see “The Tobii I-VT Fixation Filter”.

Useful Links relating to SennsLab and SennsMetrics: