Classification Of Arrhythmias Based On The GA-PNN Model

As one of the critical illnesses in the cardiovascular disease,arrhythmias will leads to sudden death or heart failure.Since therapies and treatments are different according to the kind of arrhythmias types,it is necessary to extract the information of the electrocardiogram(ECG)signals to do classification.So far,there are many researches that study the auxiliary diagnosis in this field,and the machine learning method is one of the most progressive techniques among all of the researches.The probability neural network(PNN)is a classical model in machine learning algorithms,which uses the radial basis function as the active function to handle and transmission the information.However,a smoothing parameter must be given before building the basis neural network,moreover,all of the nodes in pattern layer will use the common smoothing parameter called spread value.This thesis uses the genetic algorithm(GA)to optimize the probability neural network model,and classifies ECG signals by the optimized model.In this model,the multi-spread smoother is used to take the place of the common smoothing parameter,that is,each classes corresponding to a spread value.And this multi-spread smoother will be optimized by the genetic algorithm to improve the classification accuracy.Further more,we also did the feature process in the input layer to improve the operating efficiency.