Table 1.
Overview of several databases described in the introduction section.
Table 2.
Emotion categories used in the databases included in Table 1.
Figure 1.
Setup for motion capture sessions.
The individual in this figure has given written informed consent (as outlined in PLOS consent form) to publish the photo with the face unmasked. (A) An actor in motion capture Moven Xsens suit, t-pose (B) Acting setup: an actor in neutral pose, stool, pedals and display. (C) An actor expressing pride.
Table 3.
Stories narrated by actors during motion capture sessions.
Figure 2.
Stick-figure representations of human upper body used in the emotion categorisation studies [51].
Regardless of the body proportions of the actor, the motion trajectories were mapped onto a skeleton of average body size. Note that the motion capture files included in the database contain data for the full body and have the original actors' body size and proportions.
Figure 3.
Motion sequence frequencies across intended (A) and perceived (B, C) emotions.
Intended emotions originate from actors' text annotations while perceived emotions come from the categories forming a unique modal value in observers' response distribution for every motion sequence. The perceived emotion frequencies are split into two graphs to allow the same y-axis scale for (A) and (B) graphs. The emotional category in plot C is the sum of all frequencies in plot B.
Figure 4.
Emotion frequency distribution across consistency levels.
(A) Histogram of consistency rates across motion sequences. The minimally possible consistency is always equal to one divided by the number of observations for the given stimulus and multiplied by two because there have to be at least least two observers assigning the same category to the stimuli to form a modal value. (B) Distribution of perceived emotions across categories with a consistency rate of 0.3 or more. (C) Distribution of perceived emotions across categories with a consistency rate of 0.5 or more. (D) Distribution of perceived emotions across categories with a consistency rate of 0.7 or more.
Table 4.
Frequencies in the final set of motion sequences across intended emotion categories and acting tasks.
Table 5.
Frequencies in the final set of motion sequences across actors and acting tasks.
Table 6.
Frequencies in the final set of motion sequences across actors (rows) and intended emotion categories (columns).
Figure 5.
Emotion recognition accuracy across acting tasks for intended emotions (A) and consistency rates for perceived emotions (B) across acting tasks.
All error bars represent 95% CI.
Figure 6.
Average recognition accuracy across actors (A) and observers' response consistency across actors (B).
All error bars represent 95% CI.
Table 7.
Frequencies of motion sequences across actors where intended and perceived emotion categories coincide, sorted by the total frequency within each emotion category (rows) and actor (columns).
Figure 7.
Physical properties of motion sequences across acting tasks and individual actors.
All error bars represent 95% CI. The panels show: (A) Duration, sec; (B) Peaks in motion trajectories of right and left wrists across acting tasks; (C) Average motion speed; (D) Average motion span. As the bar plots show, physical properties of motion sequences depend both on the acting tasks and on the individual actors.
Table 8.
Online Database Overview.