Mental fatigue detection using a wearable commodity device and machine learning

Mental fatigue is a psychophysiological state that has an intense adverse effect on the quality of life, undermining both the mental and the physical health. As a consequence, detecting this state accurately can be beneficial to delivering prevention and treatment mechanisms. In parallel, advancements in wearable device technologies have reached a maturity level that can support continuous and long-term monitoring of physiological signals accurately and unobtrusively in everyday life. In this paper, a mental fatigue detection methodology is proposed, founded on the use of a wearable consumer device to enable heart rate variability (HRV) analysis and suitable machine learning models to predict the stress state with high accuracy. Even though a lot of studies have attempted to address the same problem in the past by using multiple signals, these approaches are invasive because they require a lot of sensor devices to be attached to the users. The main contributions of this work are three folds: An experimental study with 32 healthy participants demonstrating that mental fatigue caused by cognitive overload can be detected using a wearable commodity device and a single biomarker; Detection models based on eight HRV features which were found to have significant differences after inducing mental fatigue; A methodology, which includes a support vector machine, among other classifiers, and principal component analysis capable to predict cognitive performance degradation in the form of mental fatigue with a high accuracy which can be further improved by applying an ensemble model. The ability to detect mental fatigue unobtrusively and on a regular basis, for example, in workplace environments, could provide awareness on the causes of performance variations which subsequently can navigate improvements on working practices and task planning to prevent accidents or productivity losses.