| Popis: |
This paper demonstrates that an artificially generated normal electrocardiogram (ECG) using a generative adversarial network (GAN) model on the MIT-BIH arrhythmia dataset could hide the arrhythmia waveform while preserving the individual’s intrinsic characteristics. A seven-layer convolutional neural network (CNN) model was used to determine the presence of arrhythmia in normal ECG data generated through GAN, and personal authentication was performed for each ECG data. The results of each algorithm was found to be average 99.90, 99.86 percent and standard 0.18, 0.23 percent. The corresponding the F1-score was found to be 96.62, 99.13 percent and standard 0.40, 1.48 percent. |
