Research On The Detection Method Of The Mental Fatigue Based On N-back Task

With the development of society, as well as the quicken of people's pace of life and the increase of stress at work, mental fatigue appears increasingly among people who engage as military, pilots, surgeons, nightshift workers and many other groups. Serious mental fatigue may have a significant negative impact on people's work and life, and even threats human lives and property. Therefore, detection of mental fatigue is so important that it helps to provide the basis for fatigue warning, formulate the subsequent countermeasures which may eventually release or eliminate disadvantage caused by mental fatigue.This thesis designed an adjustable N-back task, instead of traditional single-recalled N-back task, we had a multiple-recall N-back task with location, color and shape comparison included. In this way, the new task could effectively induce mental fatigue by adjusting its intensity of load according to accuracy. During the experiment, electroencephalography(EEG) of subjects, fatigue scales and behavioral data of N-back task were acquired. Then mental fatigue model was built by selecting high contribution leads and features. Models in different annotations were compared. The cross-time and cross-individual rightness were tested. Because the annotation of mental fatigue is not unified, this thesis also discussed a new hybrid annotation method.This thesis had conducted several conclusions below. Using test of goodness of fit as the criterion, and search technology combined with cross-validation to select leads and features, we could build regress model of mental fatigue. Time of experiment was a better annotation method than accuracy of N-back task. Of which R2 could reach to 0.9 and EEG characteristics could well explain mental fatigue. Analysis on mental fatigue model obtained a conclusion that the model had both cross-time and cross-people rightness without sex difference. However, the model was not stable enough. Time and individual adjust were needed to improve the annotation. And, ratio of relative accuracy combined with the time of experiment could be used as an effective annotation method for mental fatigue.