ECG Signal Classification Method Based On Dynamic Fuzzy Decision Tree

ECG signal is a direct reflection of human heart health,and it is also an important basis for doctors to diagnose heart diseases.With the more and more abundant data stored in hospitals and the use of artificial intelligence algorithms in medicine,a lot of research has been done on automatic analysis and diagnosis technology based on ECG signals.But the correct diagnosis rate is low at present,which is only used as a reference for doctors' diagnosis.Therefore,it is particularly important to research the automatic analysis and diagnosis method with high accuracy.In this paper,the single cycle ECG signal is taken as the research object,and the feature extraction method and classification and recognition method of ECG signal are emphatically studied.The main research work includes the following parts:First,research on ECG feature extraction method based on time-frequency fusion feature.According to the position of R peak in ECG signal,the continuous ECG signal was divided into three layers of wavelet packet decomposition according to the period.The first four groups of wavelet packet components were reserved to reconstruct the signal and complete the denoising task.At the same tim e,the two norms of the coefficient matrix of the four groups of wavelet packet components were used as the frequency domain characteristics.Then,the information of each peak value and time interval of ECG signal were extracted as time-domain feature.Finally,the frequency domain features and time domain features were fused in the form of vectors,and the fused feature vector was used to represent the ECG signals.Second,ECG signal classification method based on dynamic fuzzy decision tree.In the process of decision tree growth,all samples of current nodes are clustered on each attribute by fuzzy C-means(FCM),which can dynamically divide the feature space,calculate the information gain before and after each attribute division,select the attribute with the greatest information gain as the split attribute.And then stop the growth of decision tree when the stop condition is satisfied.So as to construct the decision tree.The classification experiments of positive and abnormal ECG signals and multi-abnormal ECG signals were carried out using data from MIT-BIH database.The recognition accuracy reached 99.14% and 95.14% respectively.Third,design and implementation of ECG signal detection and diagnosis system.A portable ECG signal acquisition and diagnosis system was designed based on MATLAB software platform,AD8232 chip and Arduino microcontroller hardware platform.Through automatic classification of positive and abnormal ECG samples collected from 392 groups,93.88% recognition accuracy was achieve d.The experimental results in standard databases and self-collected databases show the effectiveness of the proposed method.The features extracted by the time-frequency feature fusion method in this paper can effectively represent ECG signals.The improved dynamic fuzzy decision tree algorithm increases the classification accuracy of ECG signals.It can be used as an assistant tool for diagnosis of cardiac diseases and has great importance for disease pre-detection.