The Key Technology Reasearch Of State Detection For Vigilance And Sleep

With the rapid development of brain science and artificial intelligence,the roles of human have gradually changed from workers to monitors of machinery.The monitors should retain the vigilance above a constant level.In addition,it has been confirmed that sleep plays an important role in regulating vigilance.It is an important thing to place monitors in key positions based on high sleep quality.Therefore,the research has important significance for the detection of sleep.However,there are lots of critical issues such as lacking reliability assessment for physiological signals,the poor performance for single leads selected and the analysis for sleep is two much depending on the experience and skills of professional.Aiming at these issues,this dissertation mainly focuses on the key technologies of cognitions based on the physiological signals,including the reliable of experimental dataset,the detection of vigilance state based single lead,the feature analysis and recognition of sleep.The main outcomes of this dissertation are reported as follows:(1)In the light of problem that the reliability of vigilance data,this dissertation proposed a method to verified the experimental data using statistics(t-test and effect analysis)from the perspectives of NASA-TLX scale and behavioral metrics.The proposed method can characterize the reliability evaluation of the experimental dataset.(2)In view of the problem of performance for vigilance detection and portability,we proposed a method to model the effective network based on dynamic PDC(dPDC)algorithm for band energy.The results show that the activation-related areas mainly in the right occipital regions.Recording EEG signals directly from the related regions can simple the EEG recording device and work effectively in data compression.This can also provide a theory basis for equipment miniaturization.This dissertation proposes a kernel-based ELM(k ELM)algorithm and the experimental results show that the k ELM algorithm is more effective in detecting the state of vigilance based on single lead.(3)In order to solve the problem that the CD,obtained from GP algorithm,is not efficient enough,this dissertation proposes GM-GP algorithm,which combines GP algorithm and grey model.Results show that GM-GP algorithm greatly improves the computational efficiency and can also effectively characterize the dynamic characteristics of sleep stages.The CD change trend and sleep state obtained using the GM-GP algorithm has a higher correlation coefficient.(4)This dissertation applies the CNN algorithm to the field of sleep detecting for the first time.The results show that this method has unique advantages and can effectively realize the detecting sleep state.However,it is inability to effectively recognize the awake state.Therefore,this dissertation proposes a hierarchical decision algorithm using multi-mode signals.The experimental results show that the hierarchical decision algorithm can effectively reduce the training time of the sample and ensure the accuracy of the detection.