Denoising of ECG signals in EMD and Wavelet Domains and classification with SVM Classifier

ECG signals are essential for diagnosing heart diseases but often contain noise from various sources. This study presents a hybrid denoising approach using Empirical Mode Decomposition (EMD) and Wavelet Transform (WT). EMD decomposes the signal into Intrinsic Mode Functions (IMFs), removing noise-heavy components. Simultaneously, Discrete Wavelet Transform (DWT) refines the signal by eliminating high-frequency disturbances. The cleaned ECG is reconstructed, and key time and frequency features are extracted. These features are then used to train a Support Vector Machine (SVM) classifier. The model is tested on the MIT-BIH Arrhythmia Database, achieving high accuracy. Performance is evaluated using accuracy, sensitivity, and specificity metrics. Results show that the method enhances ECG quality while preserving essential details. This approach is promising for real- time heart monitoring and automated ECG analysis.