Cardiac-Information Collection And Cognition In Remote Dynamic Electrocardiogram

As the morbidity of heart disease increases consistently, using Dynamic Electrocardiogram (ECG) to diagnose and predict heart disease has become one of the critical method in the heart disease prevention and treatment area. One drawback of ECG is that it carries a lot of information so that the combination of automatic real-time monitoring and artificial diagnostic are needed to assist ECG system. To solve this problem, we study and develop a software system which can use computer to recognize and analysis the remote dynamic ECG signals automatically.In this paper, based on the analysis of the component of mixed noise, which is introduced during signal collection and remote translation processes, we first combine Empirical Mode Decomposition (EMD) with wavelet transformation (WT) to filter high frequency noise. After pre-process, WT is also adopted to detect the location of feature points in electrical signals which are corresponding to the extreme value pair on different wavelet scales. Feature quantities that may suggest arrhythmia are then computed based on feature points. We also use these feature quantities to generate corresponding arrhythmia model through AdaBoost algorithm. Finally, the arrhythmia information automatic cognition can be realized based on the arrhythmia model we generated. The main contribution and new research finding of the paper are as followed:First, after the analysis of the component of mixed noise, which is introduced during signal collection and remote translation processes, we use WT denoising method and EMD method combined with wavelet denoising method to filter the low-frequency baseline drift noise and the high-frequency noise of industrial frequency and myoelectric interference respectively. Through the pre-process, the ECG waveform becomes clear and smooth.Second, we can achieve all the feature points of ECG by the detection of extreme values pair on different wavelet scales. This is because, there are a pair of maximal and minimal values corresponding to the ECG feature points after WT, and the location of the cross-zero point of the extreme value pair are exactly as same as the feature points. Through experiments, the robustness of this method has been confirmed. Also, the location of the signal feature points with higher accuracy can be detected.Third, in this paper, we summary and calculate14feature quantities which are used for the arrhythmia information cognitive process. We also use the feature quantities to generate corresponding different arrhythmia model through AdaBoost data mining algorithm. It has been demonstrated that using these models can realize the common arrhythmia cognition successfully.Fourth, using MATLAB and ACCESS software, we designed a relative simple system which can analyze arrhythmia automatically. It contains the signal pre-process, the detection of feature points, the arrhythmia cognition process, and the final diagnosis results. The results include basic parameters of electrical signal, effective statistical graphs, and diagnosis result of arrhythmia.