Recognition And Classification Of ECG Signals Based On Feature Extraction And Neural Network Ensemble

With the popularization of 5G wireless network,intelligent medical treatment for cardiovascular diseases with the help of Internet of things technology has become a new research hotspot.Among them,the research of intelligent electrocardiogram(ECG)wearables has attracted more and more attention.The key lies in how to realize the recognition and classification of ECG quickly and effectively without professional doctors.Therefore,this paper studies on the ECG identification and classification.Due to the weak anti-interference of ECG signals,there are a lot of noises in the collected ECG signals,which greatly affects the performance of the ECG signal classification model.For this end,the traditional ECG recognition model combines feature extraction method and machine learning algorithm to partly improve the accuracy of ECG signal recognition,but there are still some deficiencies in the stability of the model and the efficiency of training.In this connection,a series of improved methods are proposed for the preprocessing,feature extraction and classification of ECG signals.The specific work is as follows:(1)Based on IPCMM algorithm,a QRS composite wave detection algorithm is proposed,termed as II-P&T algorithm.It adopts IPCMM algorithm to generate FIR band-pass and low-pass filter to reduce the noise of ECG signal,and then combines the double slope,sliding window integral and adaptive threshold to detect QRS composite wave in ECG signal.Compared with the I-P&T algorithm and P&T algorithm in the MIT-BIH data set,and the results show that the QRS complex wave recognition accuracy of II-P&T algorithm is the highest.(2)An ensemble algorithm based on PSO-ELM(called E-PSO-ELM)is proposed.Extreme Learning Machine(ELM)has a simple model framework,which can generate parameters without iteration.It has been widely used in ECG signal classification,but ELM algorithm is not stable.To enhance the generalization performance of ELM network,particle swarm optimization(PSO)algorithm is first used to adjust the parameters of ELM to generate the PSO-ELM weak learner.On this basis,integrated multi-weak learners,an ensemble model(E-PSO-ELM)is proposed based on Bagging integration algorithm.Compared with the simulation results of other algorithms,E-PSO-ELM has good generalization performance and relatively high model stability.(3)An ensemble ELM algorithm(called as HE-ELM)based on discrete wavelet transform,linear discriminant analysis,locality preserving projections and neighborhood preserving embedding are proposed.Concretely,different features are extracted by discrete wavelet transform,linear and nonlinear manifold learning algorithm,respectively.And then,the extracted different features are integrated and combined as the input of ELM to realize the mode classification task of ECG signals.The HE-ELM simulation results show the advantages of high classification accuracy,fast running speed and high stability.