301. ECG arrhythmia Discrimination using SVM and Nonlinear and Non-stationary Decomposition
- Author
-
Yaqin Zhao, Nie Yuting, Hikmat Ullah, Fakheraldin Y. O. Abdalla, Longwen Wu, and Hassan Mkindu
- Subjects
Sample entropy ,Support vector machine ,Discrete wavelet transform ,Singular value ,Computer science ,business.industry ,Multiresolution analysis ,Statistical parameter ,Confusion matrix ,Pattern recognition ,Sensitivity (control systems) ,Artificial intelligence ,business - Abstract
ECG signals represent the all heart's electrical activity. Consequently, it performs a key function in the diagnosis of cardiac disorder and arrhythmia detection. Based on small variations in the ECG's amplitude, length, and morphology, Computer-assisted diagnosis has to turn out to be a recognized method to classifying the heartbeats of one-of-a-kind types of arrhythmia. Due to the nature of the ECG signal, a classification method was created based on the techniques of time-frequency decomposition. Discrete Wavelet Transform (DWT) was used to acquire various frequency components where Multiresolution Analysis (MRA) was implied. Based on these frequency components (MARs), the features vector was calculated using four statistical parameters. Average Power (AP), Dispersion Coefficient (CD), Sample Entropy (SE) and Singular Values (SV) were calculated from 9 RAMs as statistical parameters. SVM was then presented to use the features vector and discriminate ten distinct heartbeats of arrhythmia downloaded from the MIT-BIH database in the Physionet. Confusion matrix, Sensitivity (SEN), specificity (SPE), precision (PRE) was used and calculated to assess the efficiency of the suggested technique and compare it with the past algorithms. The performance of the suggested was discovered to be better than the current techniques, and the accuracy was 99.84 more...