Research And Implementation Of Intelligent Evaluation Method For Physiological State

The rapid rise of artificial intelligence has initiated a major transformation and ushered in a new dawn in the field of physiological status assessment,which is driven by data and intelligent status assessment algorithms.In the field of physiological state assessment,traditional abnormal heart rate states have problems of false alarms and low accuracy,while fatigue states are difficult to be sensed and assessed by a single parameter.Therefore,in this thesis,two different types of physiological data(waveform physiological signals and sparse multivariate physiological data)are used to establish deep neural network models for the recognition of abnormal heart rate and fatigue states in humans.The main research of this paper is as follows:(1)Study on the discriminative model of abnormal ECG statesIn order to solve the problem of heart rate abnormalities during exercise,this article realizes the discrimination of abnormal electrocardiogram states during exercise.Based on the electrocardiogram data in the MIT-BIH database,this article first used wavelet transform algorithm to denoise the electrocardiogram signal,followed by sample extraction based on the R peak positions annotated by experts in the dataset,and finally constructed a CNN+LSTM model to recognize five abnormal states of electrocardiogram signals.The comparative experimental results show that the new model has a 50 fold average accuracy of 99.4% for electrocardiogram signals,which is 1.13% higher than the benchmark CNN model.The new model has significantly improved its recognition ability for abnormal beats compared to CNN models,with 27,42,and4 improved recognition numbers for "ventricular premature beats(V)","left bundle branch block beats(R)",and "right bundle branch block beats(R)",respectively.(2)Fatigue state assessment model studyIn order to address the problem that soldiers’ fatigue status is difficult to be perceived in time during exercise,this thesis selected data from male subjects aged 25-35 years with good physical fitness and healthy body to simulate soldiers’ data,established a fatigue status assessment model for soldiers,and realized a comprehensive assessment of fatigue status.Firstly,physiological data acquisition experiments were conducted and physiological state labeling was performed,secondly,heart rate variability analysis was performed based on the collected ECG signals to obtain various heart rate variability indicators,and such indicators together with the relevant indicators obtained from physiological data acquisition experiments were used to construct a comprehensive fatigue state assessment index system.Finally,a deep neural network model based on recurrent neural network was established to realize the recognition of human fatigue state,and the total correct rate of the model reached 93.00%.(3)Design of physiological state assessment systemBased on C# and visual studio 2019 software platform,this thesis designs and develops a physiological state assessment system with physiological state assessment and dynamic display of physiological data.Firstly,a Bluetooth client is built based on GATT technology to acquire,parse and store physiological data.Secondly,the physiological state evaluation is realized by calling the state evaluation algorithm based on ML.NET framework.In addition,the user interface is built based on winform and SUNNYUI framework to enhance the interactivity between the system and the user.