Physiological Pattern Recognition Of Mental Fatigue Based On Electrocardiogram

Mental fatigue is a common mental state that not only damages personal physical and mental health,but also affects social stability and development.On the one hand,mental fatigue is closely related to various diseases,such as depression.Persistent mental fatigue increases the individual’s risk of illness and affects the normal functioning of the body.On the other hand,mental fatigue is a potential cause of many major accidents,posing a serious threat to industry safety.Therefore,timely and effective mental fatigue detection is crucial for early warning and intervention of physical and mental diseases and safety issues.Most of the existing fatigue studies induce fatigue based on short-term specific tasks in experimental scenarios to assess mental fatigue state.However,these fatigue inducing tasks in experimental scenarios are special cases of many fatigue triggers in real life,and whether the methods proposed in these studies are suitable for daily mental fatigue detection remains to be explored.Therefore,this work collected long-term physiological data from uncontrolled scenes in daily life,explored the autonomic neurophysiological responses of mental fatigue,constructed a physiological recognition model suitable for daily mental fatigue detection,and analyzed the mental fatigue symptoms of depressed people to obtain their mental fatigue characteristics.The specific research contents and results are as follows.(1)Collect mental fatigue data and construct physiological datasets of mental fatigue/non-fatigue and depression/non-depression.In this work,the daily electrocardiogram(ECG)data and three-dimensional acceleration data of 104 subjects were collected under uncontrolled scenarios,and the collected data were pre-processed to remove severe motion interference to obtain heartbeat interval time series suitable for mental fatigue analysis.Based on the self-assessment of mental fatigue and the two process model of sleep regulation,the mental fatigue or non-fatigue labels were marked on the heartbeat interval data,and the physiological datasets of mental fatigue and non-fatigue were constructed.The scores of Self-rating Depression Scale were used as the gold standard to construct a physiological dataset of depression and non-depression.(2)Explore the autonomic nervous response of mental fatigue,construct a physiological recognition model of mental fatigue,and explore the characteristics of mental fatigue in depressed people.This work applied t-test and Mann-Whitney U test to analyze the impact of mental fatigue on heart rate variability parameters,and the results showed that there were significant differences between mental fatigue and non-fatigue on multiple heart rate variability parameters.A variety of supervised machine learning classifiers were used to train the mental fatigue physiological recognition model,and the generalization performance of the model was evaluated by an independent test set.The results showed that the back-propagation neural network classifier obtained the most balanced performance,achieving 87.80% and 89.42% F1 scores of fatigue recognition in the validation and testing processes,respectively.Based on the above mental fatigue recognition model,mental fatigue detection was conducted from 9 a.m.to 9 p.m.in the depressed and non-depressed groups,and the detection rates of mental fatigue in the two groups were counted at half-an-hour intervals.The results showed that the proportion of mental fatigue detected in depressed and non-depressed populations appeared a similar downward trend in 09:30-11:30 and 14:30-16:30,but in the non-depressed group,the downward trend of the proportion of mental fatigue continued for a longer time in the afternoon and reached the lowest value at 16:30-17:00.In many time periods,the detection rate of mental fatigue in the depressed group is higher than that in the non-depressed group,especially in 09:00-09:30,12:30-13:00,14:00-14:30,and 16:30-18:00.In 13:00-13:30 and 18:30-19:00,the detection rate of mental fatigue in the non-depressed group increased,while the detection rate of mental fatigue in the depressed group decreased,reaching its lowest point in 18:30-19:00.The research findings of this work are as follows.(1)Mental fatigue had autonomic neurophysiological responses different from non-fatigue.Compared with the non-fatigue state,the sympathetic nervous activity increased and the parasympathetic nervous activity decreased in the mental fatigue state.(2)Mental fatigue and non-fatigue had distinguishable neurophysiological patterns.The accuracy of the mental fatigue recognition model in this work was much higher than that of random guess,which can effectively identify the mental fatigue state of individuals.(3)There were both similarities and significant differences in the performance of mental fatigue between depressed and non-depressed people.The similarity was that the mental fatigue state of both depressed and non-depressed people showed a decreasing tendency within certain time period.The differences were mainly reflected in the following two aspects: First,the depressed group had more mental fatigue state than the non-depressed group for many periods of time.Secondly,depressed and non-depressed people showed opposite characteristics of mental fatigue in some time periods,i.e.when the degree of mental fatigue in the non-depressed group increased,the degree of mental fatigue in the depressed group decreased.