Datasets and Analyses for "Affect Recognition using Psychophysiological Correlates in High Intensity VR Exergaming"

Data collection method:

We used a Lode Excalibur Sport exercise bike and an FOVE HMD. They were connected to a PC running Unity with an Intel Xeon E5 2680 processor, 64 gigabytes of RAM, and two NVIDIA Titan X graphics cards. We measured blink rate in blinks per minute (Blinks) with the eye gaze tracker built into the FOVE HMD, recording pupillometry data with FOVE’s Unity plugin at 160Hz and counting blinks as periods with zero pupil diameter. We measured the tonic skin conductance (Conductivity) using the Shimmer3 Consensys GSR development kit in microsiemens (μS) at 128 Hz. Furthermore, we recorded the average power output (Power) in Watts during the sprint phases in conditions. Experiment I: We collected ground truth data for affect based on validated post-condition questionnaires. We measured intrinsic motivation with the Intrinsic Motivation Inventory (IMI) . We used the main Interest/Enjoyment subscale (IMI Enjoy) with a scoring ranges from 1 to 7, with 7 being the highest intrinsic motivation score. Participants then performed each of the four conditions: B (Baseline), G (Game), E (Exercise) and EG (Exergame). After conditions G, E and EG, participants completed the IMI, and left qualitative feedback about their experience. Experiment II: In addition to recording Conductivity, Blinks and Power to determine affective state, we recorded the total time of eye gaze fixations (Fixations) on visual components of the game: the competitor, the gap between the player and the competitor, the points, prompts, the displayed RPM and the timer. We used ray casting to detect the game components corresponding to a point of gaze. A low Fixations value indicates that the player was looking more at the peripheral VR environment or ‘staring at nothing’ instead of paying attention to the game. We also recorded a participant’s pupil dilation (Pupil) during the warm up and in each of the two sprints, considering their average. Similar to Experiment I, we used the IMI Interest/Enjoyment subscale (IMI Enjoy) to measure intrinsic motivation. Lastly, the experience sampling method integrated in the exergame was used to collect ground truth values about the player’s affective state; we consider the average of all values measured in a condition (Affect). We matched the sensor data and the ground truth by taking the average of the sensor data and of the experience sampling measures over a whole gameplay session.

Technical details and requirements:

We used a Lode Excalibur Sport exercise bike and an FOVE HMD. They were connected to a PC running Unity with an Intel Xeon E5 2680 processor, 64 gigabytes of RAM, and two NVIDIA Titan X graphics cards. We measured blink rate in blinks per minute (Blinks) with the eye gaze tracker built into the FOVE HMD, recording pupillometry data with FOVE’s Unity plugin at 160Hz and counting blinks as periods with zero pupil diameter. We measured the tonic skin conductance (Conductivity) using the Shimmer3 Consensys GSR development kit in microsiemens (μS) at 128 Hz. Furthermore, we recorded the average power output (Power) in Watts during the sprint phases in conditions. For the II Experiment, in addition to recording Conductivity, Blinks and Power to determine affective state, we recorded the total time of eye gaze fixations (Fixations) on visual components of the game: the competitor, the gap between the player and the competitor, the points, prompts, the displayed RPM and the timer. JASP statistics software (https://jasp-stats.org/) and the R programming language were used for data analysis. For the R scripts, we recommend using the RStudio IDE (https://rstudio.com/).

Methodology link: