Denoising and Time-Frequency Features Extraction of ECG Signals Using Iterative Unbiased FIR Algorithm

Cardiovascular diseases are known to be leading causes of death in our life. Electrocardiogram (ECG) recordings are crucial for diagnosing heart-related conditions because they capture important morphological characteristics useful for automatic pathology detection systems. However, these characteristics can be affected by noise or artifacts. While several techniques have been developed to address this issue over decades, there is still a need to improve the accuracy of automatic ECG signal detection and classification. In this study, we propose using robust unbiased finite impulse response (UFIR) technique for denoising and feature extraction of ECG signals showing arrhythmias, congestive heart failure, and normal rhythm. The UFIR filter superior performance is compared to other filters in terms of the root mean square error (RMSE) under different noise levels. Testing is conducted using ANOVA and Kruskal-Wallis analysis of time-frequency domain features, focusing on the spectral kurtosis of the UFIR filter. It is shown that the UFIR filter provides features with highly significant differences between the pathologies.