Research On Pulse Signal Analysis And Recognition Based On Deep Learning And Ensemble Learning

Pulse diagnosis is an important part of Traditional Chinese Medicine(TCM),which is indispensable in clinical diagnosis.TCM doctors diagnose patients by subjective feelings and long-term experience,lacking objective evaluation standard.In the view of above issues,this paper applies a variety of signal analysis methods and machine learning algorithms to analyse and recognize pulse signals,which lays a certain foundation for the objectification and intellectualization of pulse diagnosis.The main content of this paper can be divided into the following three parts:Firstly,this paper used Lyapunov exponent to analyse the nonlinearity of pulse signals,which showed that pulse signals have chaotic characteristics.On this basic,the pulse signals were converted into non-threshold recurrence plots by nonlinear dynamics theory,avoiding a large number of details loss due to a wrong selection of threshold.The convolutional neural network was used to extract the nonlinear features of the recurrence plots,and established a classification model of pulse signals automatically.The experimental results show the proposed method is effective.Secondly,in order to make better use of the advantages of different pulse signal analysis methods,this paper proposed an integrated classification model named ResNet and SVM based Stacking Networks(RSSN),combining time domain,time-frequency domain and nonlinear dynamics analysis methods.SVM was used to establish classification models for features extracted by time and time-frequency domain analysis methods;ResNet was used to establish classification model for non-threshold recurrence plot.Using Stacking method to integrate SVM and ResNet can combine the advantages of different models to some extent.The experimental results show that the RSSN algorithm can effectively improve the accuracy of pulse signals recognition and has excellent classification performance.Thirdly,this paper designed and developed a TCM pulse diagnosis collection and analysis system,which used Visual Studio 2015 development tools and DuiVision interface library.This system realized the the collection of patients' basic information and pulse signals,and then analysed and classified the collected pulse signals by extracted features.Based on the eight factors of pulse,this paper proposed a classification framework for 29 pulse types by threshold method and RSSN,and established classification models for some factors based on existing datasets.Finally,the MySQL database was used to implement data storage and query functions.