Abstract：Objective In this paper, we proposed a patient-specific electrocardiograms (ECG) classification method based on onedimensional
convolution neural network to improve the performance about the automatic classification of heart beat, especially the
supraventricular ectopic beats. It would provide an auxiliary basis for clinical ECG diagnosis. Methods We combined the ECG
features of the multi-layer one-dimensional convolution neural network and the RR interval characteristics of ECG, and then they
were sent them into the multi-layer sensor. After that we classified the features processed by softmax classifier. In order to achieve
a better patient-specific heart beat recognition, we added the specific-patient ECG data into the original part of the model training
data from the public data set. Results Compared with the existing research results, the performance of classification using the MITBIH
arrhythmia database was improved. The sensitivity of SVEB recognition was increased from 60.3% to 88.7%. Conclusion This
method can provide a reliable basis for the diagnosis of heart disease.
