Research On Classification And Recognition System Of ECG Signal Based On RBF Neural Network

Electrocardiogram(ECG)is a bio-electrical signal produced by cardiac activity,which can directly reflect the health of the heart.there is a great significance in clinical practice of use ECG signals to analyze and diagnose cardiovascular diseases.Therefore,the classification and recognition technology of ECG signals has become a hot topic for scholars both at home and abroad.This paper mainly studies three aspects of perprocessing,feature extraction and classification recognition of ECG signal classification and recognition system.In order to improve the accuracy of classification and recognition of ECG signals,the following work is done in this paper.(1)As the ECG signal is a weak electrical signal,it is extremely vulnerable to interference.the common noise includes myoelectric interference,power frequency interference,baseline drift,etc.In this paper,according to the characteristics of several noises,FIR filter,Butterworth filter and median filter are used to filter out the noise interference in ECG signals.Through the simulation results,it is found that the noise interference is removed and the characteristics of the electrical signal can be well matained.(2)Due to defects of the traditional differential threshold method,this paper proposes a R wave peak location method based on moving window,first order difference and two order differential square.And,on this basis,the single QRS group can be obtained by intercepting 62 data points on both sides of the R wave peak.(3)Because of the high compression ratio and fast compression speed,the wavelet transform can keep the characteristics of the signal and the image unchanged.In this paper,Harr,db2 and bior2.4 wavelet transform of single QRS wave group are carried out respectively.By comparison of experiments,this paper uses bior2.4 wavelet to extract the feature of ECG signal.(4)The BP neural network and RBF neural network are used to classify and recognize the ECG signals,and the RBF neural network is selected for classification and recognition of ECG signals through experimental comparison.Then as the traditional RBF neural networks artificially determine the number of hidden nodes and the center of radial basis funcion,this paper uses the sub-clustering algorithm to optimize the structure of the RBF neural network and proposes the selection of training samples by the distance from the cluster center.The simulation results show that the improved neural network learning algorithm can obvisously improve the recognition rate of normal cardiac beat,left bundle branch block,right bundle branch block and ventricular premature beat compared with the traditional randomly selected training samples.It shows that this network has better performance in ECG classification and recognition.Lastly,this paper uses C# language based on the Visual Studio 2013 platform to complete the writing of ECG signal classification and recogition system.