Research On ECG Signal Classification Based On Deep Learnin

ECG is a physiological signal that characterizes the state of the human body,and the identification of ECG is crucial for the diagnosis and treatment of CVD.However,ECG is extremely sensitive to noise when collected,and the large number of ECG data types and unbalanced distribution of ECG samples can lead to unsatisfactory classification performance.To address these problems,the main research of this paper is as follows.(1)Considering the noise sources and characteristics of ECG,this paper proposes a new noise reduction algorithm for ECG by combining Butterworth filtering with median filtering.Firstly,the Butterworth filter is used to suppress the high frequency noise.The Butterworth filter can smoothly suppress the high-frequency noise while retaining the detailed features of the original signal.Secondly,the median filtering method is found to have better noise reduction effect based on low-pass filtering,so the median filtering is used to filter the signal processed by Butterworth filter to complete the whole denoising process.(2)Due to the obvious data imbalance problem among various types of ECG samples,a single generative model cannot guarantee a good classification of the generated heartbeats.In this paper,we propose the GAN-Tomek Links method to equalize the segmented ECG data.Firstly,we use the trained generative model and discriminative model to form a generative adversarial network for ECG signal generation.Secondly,the Tomek Links undersampling method is used to process the generated overlapping samples and strengthen the generated data.(3)In order to better extract the features of ECG signals,multiple modified SE modules are embedded in the CNN to extract the key features of ECG signals from the perspective of channels.BLSTM to extract the ECG time series information and build up several branches of the model to extract ECG data independently.At the same time,this paper compares with other research methods based on two cases before and after data balancing,and the results show that the proposed ECG signal classification method has higher classification performance.