Effective Hybrid Technique for EcG Signal Denoising Based on Slantlet Transform and Median Filter

Denoising techniques based on wavelet transform (WT) and empirical mode decomposition (EMD) have been used widely to remove different types of noise. However, these techniques have limitations regarding edge effects. This work aims to denoise the ECG signals and preserve ECG's crucial characteristics using Slantlet Transform (SLT) and Median Filter (MF). 48 recordings of ECG signals from Physionet's MIT-BIH arrhythmia database were tested by adding different levels of Gaussian white noise (AWGN). 10-SLT filterbanks were used to extract the different frequency components using the decomposition process. An adaptive thresholding approach was used to calculate a suitable threshold for the amount of noise to achieve a minimum mean square error (MSE) value. Hard and soft thresholding methods were used to denoise the generated SLT coefficients. SLT performance was enhanced by passing the reconstructed clean coefficients to adaptive MF for removing residual noise and maintaining the important ECG features. The average results of ECG records showed that the improving signal-to-noise-ratio (SNRimp) was 8.098-dB at 10-dB which represents a significant improvement (2.41-44^%) compared to EMD and its improved versions, Sparsity Assisted Signal Smoothing, and Non-local Means. At 5-dB, SNRimp, percent root mean square difference (PRD), and MSE were 9.7481dB, 18.4426%, and 0.0153, respectively; to achieve improvements up to 6.39%, 6.47%, and 18.97%, respectively; compared to adaptive dual threshold filter and the discrete wavelet transform. In addition to reducing noise, the proposed method kept the ECG characteristics. This method has shown considerable improvements in SNRimp, MSE, and PRD compared to other methods.