Signal Multimodal Physiological Fusion for Cognitive Workload Classification in Intelligent Immersive VR-Based Geriatric Nursing Training
Description
This dataset contains raw, time-synchronized multimodal physiological recordings from six participants performing VR-based geriatric nursing training. It was collected to support assessment and classification of cognitive workload from fused physiological signals. All signals are provided in their original acquisition units with no filtering, artifact rejection, normalization, or feature extraction applied. Records are organized on a common 250 Hz timebase (4 ms resolution), with the Time column giving seconds elapsed from the start of the recording. Signals EEG — 14 channels (AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4, F8, AF4), in a montage consistent with a 14-channel wireless headset (EPOC X). Values are unreferenced and retain the full DC offset, typically in the 4,200–4,950 range; users should apply appropriate referencing and high-pass filtering before analysis. Eye tracking — 3D gaze direction vector (GazeDirection_X, _Y, _Z), identifier of the object under gaze (LookedObject), current fixation duration (GazeDuration), cumulative gaze time on the current object (TotalGazeTime), and pupil dilation for the left eye, right eye, and their average, in millimeters. Cardiovascular — peripheral oxygen saturation (SpO2(%)) and pulse rate (Pulse Rate(bpm)). The disparity is largest for the cardiovascular channels. SpO₂ and pulse rate were recorded at approximately one sample per 1,000 EEG samples — roughly 0.25 Hz, or one reading every four seconds. Task context — Module, an integer index identifying five different modules of the training task. Label — Cognitive Load, a categorical workload annotation with levels [low/ Medium/ High ], derived from Cognitive Load Scale survey completed after each module was completed.
Files
Institutions
- Mississippi State UniversityMississippi, Starkville