A review of Convolution Neural Network, ECG and STFT Based Spectrogram for Health Applications

  • B. V. Durga, S. B. S. Lakshmi, S. Musthafa, B. Kalesh, S. Kumar

Abstract

The Electrocardiogram (ECG) is very useful to diagnose heart diseases. It has two steps, feature extraction, and pattern classification. The abysmal neural network that was trained on enormous information can carry out feature extraction from records, and also diagnosed the cardiac arrhythmia healthier than a specialized cardiologist. For ECG arrhythmia arrangement the deep 2-dimensional convolution neural system is used. The time-domain five different types of ECG signals are converted into a frequency spectrogram by using an STFT. The five ECG spectrograms are given as input to the 2-D convolution neural network so that ECG arrhythmia signals are classified. The proposed 2-D CNN method achieves an average accuracy of 99.00% and also having optimal classification performance at the education rate is 0.001 and the bunch size is 2500. Without manual pre-processing, it achieves more accuracy and low losses than the 1-D CNN model. So, the 2-D CNN method is more widely used nowadays.

Published
2020-12-01
How to Cite
B. V. Durga, S. B. S. Lakshmi, S. Musthafa, B. Kalesh, S. Kumar. (2020). A review of Convolution Neural Network, ECG and STFT Based Spectrogram for Health Applications. International Journal of Advanced Science and Technology, 29(04), 11197 - 11206. Retrieved from http://sersc.org/journals/index.php/IJAST/article/view/34443