Research On Sleep Disorder Recognition Methods And Key Technologies Based On Physiological Information

Insomnia is a typical sleep disorder.Insomnia will not threaten people’s life safety in a short time,if the disease is allowed to develop,it will bring a series of concurrent diseases and even lead to death.The most effective way to reduce the risk of insomnia is early diagnosis and early treatment.However,the traditional artificial method lack of effective physiological indicators for automatic recognition,which are time-consuming and laborious,and easily affected by subjective factors of doctors.To solve this problem,this paper takes automatic recognition of sleep disordersas starting point for research.In turn,the screening of sleep physiological features,automatic sleep staging,effective recognition of insomnia were studied.The main work and contributions are as follows:1.In order to explore the valuable automatic insomnia recognition information contained in the sleep data,a physiological data feature selection algorithm based on redundant removal was proposed.The vertical correlation(the relationship between features and class attributes)and the horizontal correlation(the relationship between features and features)are analyzed to determine two redundancy criteria.To quantify redundancy criteria,an approximate redundancy feature framework based on mutual information is defined to remove redundancy and irrelevant features.To evaluate the effectiveness of the proposed algorithm,the validation is carried out on eight public biological datasets and compared with the typical feature selection algorithm.The experimental results show that the proposed algorithm can effectively reduce the feature dimension and improve the classification accuracy.In addition,experiments show that the optimalvalues of the core parameters δ and α are 0.05 to 0.13,and 0.60 to 0.66.2.Explore the representation and management methods of non-structural sleep physiological features,and automatic sleep staging,an automatic sleep staging scheme based on ontology and feature weight analysis was designed and implemented.After preprocessing the physiological data,an ontology model is constructed to store and manage a large number of physiological features and other sleep context information.Based on weighted feature analysis,an improved random forest classification algorithm is proposed to realize automatic sleep staging.Experimental data analysis results show that the average accuracy of the five-state sleep stages of the improved random forest algorithm improved by 5.00%.Meanwhile,experimental results also show that the improved random forest algorithm is less affected by the number of sleep segments.3.On the basis of the first two works,an automatic insomnia recognition method based on universal physiological data and graph theory is proposed.Firstly,a universal sleep experiment based on the portable physiological data acquisition instrument developed by our research group is designed.Secondly,the optimal feature subset of insomnia recognition is selected based on the feature selection algorithm of physiological data with redundancy removal.Thirdly,the combination learning method is used to explore the optimal sleep staging stage of insomnia recognition.It mainly includes three aspects:combination learning to obtain the optimal sleep feature subset of insomnia recognition,combination learning to obtain the optimal classifier of insomnia recognition,and combination learning to discover the optimal sleep stages of insomnia recognition.Then,the optimal feature subset is mapped to the graph node,and thestructure graph is constructed based on Euclidean distance.Finally,insomnia recognition based on optimal feature subset and graph attributes as input of classifier.Experiments show that:(1)The best sleep stage of insomnia recognition,male is SWS stage,and female is REM.(2)The most effectivefeatures of male and female insomnia recognition are screened out by feature selection algorithm based on redundant removal of physiological data.(3)Structure graph and improved random forest algorithm are used to realize automatic insomnia recognition.The recognition rates of male and female are 93.94% and 95.13%,respectively.In summary,this paper has carried out exploration and research on physiological information modeling methods and key technologies for automatic recognition of sleep disorders,and focuses on key problems of sleep physiological features screening,automatic sleep staging,and effective recognition of insomnia and so on.The corresponding solutions are proposed,and the effectiveness is verified by experiments.The feasibility of the new technology and new theory is objectively discussed from multiple angles.The work done in this paper has promoted further research and development of automatic recognition of sleep disorders.