Fatigue is a multidimensional construct with experiential (e.g., feelings of tiredness), behavioral (e.g., performance deficits), and neurophysiological (e.g., increased EEG alpha and theta wave activity) dimensions. Modeling mental fatigue based on physiological signals is challenging due to a lack of understanding of the interplay between physical activity and mental performance, as well as the difficulty of measuring fatigue in naturalistic settings.
We present a dataset containing multimodal sensor data from four wearable sensors during controlled physical activity sessions. Researchers can use this data to characterize the effect of physical activity on mental fatigue, and to predict mental fatigue and fatigability using wearable devices.
FatigueSet contains approximately thirteen hours of sensor data collected over 36 sessions (twelve each of low/medium/high activity). Data were recorded from 14 sensors on four wearable devices at different on-body locations:
|Nokia Bell Labs earable prototype||Accelerometer|
|Muse S EEG Headband||Electroencephalography|
|Zephyr BioHarness 3.0 chest band||Electrocardiography|
|Empatica E4 wristband||Accelerometer|
|Blood volume pulse and IBI|
Additionally, FatigueSet also contains subjective fatigue ratings and mental fatigability scores on two cognitive tasks - the Choice Reaction Time task and the 2-back task. For more details, please see the below publication.
The dataset is structured as below:
|━┗ <data files>|
Each participant directory contains three subdirectories corresponding to each session. The format of data files in each session folder are described in README.md in the main folder. This file also describes the metadata and survey response files in the main folder.