An integrated EMD adaptive threshold denoising method for reduction of noise in ECG

Abstract

Electrocardiogram (ECG) denoising is a biomedical research area of great importance. In this paper, an integrated empirical mode decomposition adaptive threshold denoising method (IEMD-ATD) is proposed for processing ECGs. Three methods are included in the IEMD-ATD. First, an integrated EMD method based on a framework of complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is proposed to improve the decomposition quality and stability of raw ECGs. Second, a new grouping method for intrinsic mode functions (IMFs) is developed based on the energy and eigenperiod of IMFs. The grouping method is able to determine the boundaries among high-frequency noise predominant IMFs, useful information predominant IMFs and IMFs with low-frequency noises. Finally, an adaptive threshold denoising method is derived and used for denoising high-frequency noise predominant IMFs. There are two main contributions: 1) an adaptive threshold determination method based on the 3σ criterion and 2) a peak filtering denoising method for retaining useful information contained in the values smaller than the threshold. Synthetic and real ECG data in the MIT-BIH database are utilised in experiments to illustrate the effectiveness of IEMD-ATD for ECG denoising. The results indicate that IEMD-ATD offers better performance in improving the signal-to-noise ratio (SNR) and correlation coefficient compared with the existing EMD denoising methods. Our method offers obvious advantages, especially in retaining detailed information on the QRS complex of the ECG, which is significant for the feature extraction of ECG signals and for pathological diagnosis.