Heartbeat Pattern Analysis And Abnormity Detection Under Accumulative Effect Of Stress

Stress refers to physical and mental tension of people resulted from environmental stimuli.Moderate stress can stimulate physical arousal,allow people to better respond to external environmental stimuli and enhance body adaptability.However,long-term stress can easily lead to fatigue,chronic inflammation and metabolic syndrome,destroy the immune system function and induce mental and physical diseases.At present,most stress physiology pattern recognition studies focus on the analysis of the heartbeat pattern of short-term acute stress events.There is no literature that proposes effective quantitative indicators and physiological computational models of the accumulative effect of sustained stress on autonomic nervous system.This thesis assumes that the recurring acute stress events have an accumulative effect.Strong and weak stress elicitation tasks were designed,each lasting for about 8 hours in two real-world scenarios.We used the multi-scale heartbeat pattern analysis technique to verify the accumulative effect of stress over time.Data analysis showed that the accumulative effect of stress reduced heartbeat complexity at multiple scales,and the range of local Hurst exponents(RLHE)effectively indicated the development process of the accumulative effect of stress.The main research contents and results are as follows:(1)We designed an experimental paradigm of continuous strong/weak stress elicitation and data acquisition.Through about 8 hours of real-world computer game tasks and computer learning tasks,we induced continuous strong stress of 30 subjects and continuous weak stress of another 30 subjects.Supervised physiological data collection were performed during stress elicitation tasks,and the stress labels of the physiological data were comprehensively calibrated by using the subjective stress information and objective physiological changes.The statistical test results showed that strong/weak stress elicitation tasks successfully induced continuous strong/weak stress status of the subjects.(2)The running window method was applied to construct a large heartbeat dataset of strong/weak stress status under observation windows with different window lengths.Using the large dataset,we analyzed the ability of four heartbeat indicators,i.e.heart rate running mean(RM),average small-scale heart rate fluctuation(AFSS),average large-scale heart rate fluctuation(AFLS)and RLHE,in distinguishing strong and weak stress at different scales.It was found that the differences of RLHE index between strong and weak stress increased with the increasing observation scale and window length,indicating that the accumulative effect of sustained stress did exist.The accumulated stress leaded to more significant complexity reduction of heartbeat at large scales.(3)We used RM,AFSS,AFLS and RLHE to construct the feature space,and the heartbeat data distributed in the feature space were classified into strong or weak stress subsets by four common classifiers.Using 2-foldrandom verification method,the K-nearest neighbor(KNN)classifier had a strong/weak stress recognition accuracy of 74.35%(F1 score)at the observation window of 104 heartbeats,and the strong/weak stress recognition accuracy of KNN at the observation window of 4800 heartbeats increased to 99.63%.KNN was the classification method with the most improved accuracy and the highest accuracy among the four classifiers.The scatter distribution pattern of data samples in the feature space indicated that the KNN pattern recognition method most adequately captured the sub-region centralization trend of the strong/weak stress data caused by the accumulative effect of stress.(4)The data of the subjects whose heartbeat complexity was extremely reduced at large scale were analyzed.We found that the extreme reduction of heartbeat complexity at large scale only occurred in the subjects with extreme fatigue caused by sustained strong stress,indicating that the extremely low heartbeat complexity at large scale was a strong evidence of stress-induced autonomic fatigue.Therefore,the RLHE index provided an objective quantitative indicator for early warning of heart accidents.