Analysis of ECG De-Noising Using Non-Local Means with Approximate Coefficients and Particle Swarm Optimization

Electrocardiogram (ECG) is a widely employed tool for clinical diagnosis. Removing ECG noise to obtain a high-quality ECG signal is the premise of subsequent analysis. In this study, we proposed a robust hybrid method for removing Additive White Gaussian Noise (AWGN) from ECG signals. Firstly, AWGN was artificially added to clean test signals from MIT-BIH database for single-scale Discrete Wavelet Transform (DWT) decomposition to obtain approximate coefficients, which are then reconstructed for Non-Local Means (NLM) estimate. A Linear Decay Weighted-particle swarm optimization (LD-PSO) technique is used to search the key parameters in the NLM method. Finally, the denoising process is realized by re-modeling according to the searched coordinates. The proposed method was tested in the MIT-BIH arrhythmia database and evaluated by mean square error (MSE), SNR improvement (SNRimp), and distortion percentage (PRD). The effectiveness of the proposed method is validated by comparing it with evaluation metrics of the existing similar methods. The proposed algorithm effectively removes the AWGN from ECG signals with less manual intervention. This work might help to obtain better quality ECG waveforms for diagnosis.