A Study Of QRS Wave Detection And Classification Methods For ECG Signals

Cardiovascular disease has become one of the main diseases that endanger human health.In order to help doctors accurately diagnose the disease,electrocardiogram has become one of the most commonly used and most important examination measures,The ECG is widely used in medicine by recording the corresponding ECG waveform generated by each heartbeat action cycle.The main functions include:monitoring the electrophysiological activities of normal myocardium,normal physical examination and so on.Different waveforms of ECG signals can reflect different types of lesions in the heart,and identifying different types of ECG waveforms is more helpful for doctors to make a correct clinical diagnosis.However,even experienced cardiologists cannot directly make an accurate diagnosis only by recognizing the electrocardiogram,and often require other examination methods such as echocardiography to make a definite diagnosis.In this process,there will be serious situations such as excessive extension of the diagnosis process and misdiagnosis.According to the importance of ECG signals in the whole diagnosis and treatment process and the problems that may occur in the current diagnosis and treatment process due to the excessive extension of the diagnosis process and misdiagnosis,which may affect the quality of life of patients due to missing the best diagnosis and treatment time,there is an urgent need for a solution that can A method to quickly assist doctors to make a clear diagnosis and carry out subsequent related treatment.As a new type of medical aid,the automatic detection and classification algorithm of ECG signal can effectively detect the most informative QRS wave in the ECG signal,and can classify and diagnose the ECG signal waveform corresponding to different types of lesions,which greatly solves the problems mentioned in the current diagnosis and treatment process mentioned above.Based on this,the research content of this paper is mainly divided into two parts,one part is the detection of the QRS wave in the ECG signal,and the other part is the classification of the ECG signal,as follows:(1)In order to reduce the influence of noise in the process of QRS signal recognition and detection,a new method based on mathematical morphology is proposed.Because the ECG signal is very weak and noise is inevitably introduced in the process of QRS acquisition,in order to reduce the influence of noise,the ECG signal must be filtered.Compared with the FIR filter,the method of mathematical morphology is not only simple in calculation,but also less in steps and more effective in removing the noise in ECG signal.In QRS detection,in order to improve the accuracy of QRS group detection and recognition,this paper uses adaptive dynamic double threshold method to identify and locate QRS wave.The adaptive threshold can change according to the peak value of R wave,which can improve the detection accuracy effectively,and the double threshold can reduce the probability of miss detection and error detection.Through the simulation on MATLAB platform,it can be seen that the proposed method can achieve high recognition accuracy for both ECG and measured signals in arrhythmia database.(2)In order to improve the accuracy of ECG classification using convolution neural network algorithm,this paper proposes a convolutional neural network ECG signal classification model bas ed on wavelet transform.Firstly,the QRS wave group is located,and a certain number of sampling points are intercepted to form a new beat signal as the input signal of the subsequent ECG signal classification model.After that,this paper uses discrete Dobbesian wavelet to decompose the intercepted heartbeat signal into feature coefficients,and further reduce the dimension of the feature coefficients matrix according to the combination of the detail coefficients and approximate coefficients.After this processing,the input signal reduces the dimension of the input signal compared with before processing because some redundant feature information is dropped.The network model trained with this data has better generalization ability and can train the network more quickly.The network model has higher classification accuracy when tested with test set.In order to verify the superiority of the proposed algorithm,the SVM algorithm,convolution neural network algorithm and convolution neural network model based on wavelet transform are used to classify 48 ECG signals.The simulation results show that the proposed algorithm can not only improve the accuracy of SVM algorithm and convolution neural network algorithm,but also reduce the training time of the model.