ECG Singal Automatic Recognitioin Technology

Electrocardiogram (ECG) is clinically applied widely as it provides rich physiologicaland pathological information which offer key insights into the basic condition of heart.Efficient diagnosis of ECG heartbeats is fundamental as well as essential for clinical therapy.This paper explored the ECG signal automatic recognition technology on the basis ofdomestic and international research and predecessor’s work.The automatic recognition technology mainly consists of signal collecting, preprocessingunit, feature acquisition and classification. The preprocessing unit is proved to beindispensable. From the Matlab experimental results, both of the effectiveness and usefulnessstrategy for preprocessing is distinctively confirmed.The study is engaged in building a two-lead ECG feature model strategy to prompt thecurrent single-lead method. The popular method only uses the information of a single leadafter locating the main points of the single with two-lead information. For taking the mostdiscriminative feature of the ECG signal, both of the recording leads are employed in thepaper. Independent Component Analysis (ICA) is utilized for feature extracting, theperformance of which is highly decided by the goal function and the optimization method.The theory of ICA is introduced. The FastICA with negentropy as goal function is detailedand used as the final implementation of ICA.The classification is the last and key part of the recognition. Support vector machine(SVM) is a widely studied algorithm. Different kernels and the feature modules lead todifferent classification performance of SVM classifier. The most popular kernel functions arelinear kernel function, polynomial kernel function and Gaussian kernel function. The last oneis proved to be the most appropriate one for ECG signal feature model. q Gaussian function isa generalized Gaussian function in nonextensive statistic area. Focusing on the relation of theq Gaussian function and the Gaussian function, q Gaussian function is implemented to theSVM as a kernel function and adopted to classify the ECG signal in the paper. The results alsoindicate the good effect of the idea. What’s more, all the contrast experiments present that thetwo-lead ECG feature model strategy is better than the single-lead method.The paper improves the ECG signal automatic recognition technology in the featuremodel strategy and classification. Two-lead ECG feature model strategy is implemented in thefeature acquisition part. Meanwhile, the q Gaussian kernel SVM is constructed and applied inECG signal classification. With the standard MIT/BIH arrhythmia database, simulationexperiments are carried out. Both of the ideas are validated to be effective.