Feature extraction of electrocardiogram signals based on Wavelet Transform and Modulated Sine Modulated Filter Banks

Document Type : Original Article

Author

Alexandria university

Abstract

Cardiovascular diseases are the common cause of death among several diseases. Electrocardiogram (ECG) is widely used to test the efficacy of the heart of patients. Analysis of ECG signals using computer techniques is usually applied to help the physicians in their diagnoses’ decisions. Many researches are conducted to obtain the feature extraction of ECG signals. In this paper, several trials are compared to obtain the maximum accuracy and maximum area under curve (AUC). New approach of filter banks is presented in this paper, which is called modulated sine modulated filter banks (MSMFBs) based on wavelet transform, using deep learning as a classification method. Wavelet transform is calculated for the input ECG signals to obtain 8 coefficients for each. Then, MSMFBs is obtained for each coefficient. After that, minimum is computed. Finally, deep learning (RNN based BILSTM) is used in the classification phase. The obtained accuracy rate is 99.14%. The maximum area under curve (AUC) is 0.9682.

Keywords

Main Subjects

Files

Share

How to cite

Statistics