Noise and artifact reduction based on EEMD algorithm for ECG with muscle noises, electrode motions, and baseline drifts

As the number of ageing population and cardiovascular diseases increased, innovations in electrocardiogram (ECG) recording devices with more compact design and capability to be operated while user moving freely and comfortably are needed to help monitoring cardiac activity. Despite of its advantages, the high mobility of ECG leads to the increasing of motion artifacts and baseline drifts which then become another challenge to be overcome. A method based on Ensemble Empirical Mode Decomposition (EEMD) algorithm is proposed in this research for reducing noises and artifacts caused by motion, e.g. muscle movement, baseline drifts, and electrode motions. Testing was done by generating noisy signal using three types of noise recordings (taken from MIT-BIH Noise Stress database) and normal ECG recordings (taken from MIT-BIH Arrhythmia database with normal annotations), which each type of noisy signals divided into five noise levels i.e. 2, 6, 10, 14, and 18 dB. The performance of proposed method was then evaluated qualitatively by asking opinion from qualified general practitioners and quantitatively by calculating the SNR values. The output signals are then compared with output from Finite Impulse Response (FIR) filter and Empirical Mode Decomposition (EMD). As the results, for three types of noise tested, the ECGs are qualitatively tolerable for being used for diagnosis if the SNR are equal or above 9 dB. According to that, proposed method is effective for reducing muscle noise and electrode motion artifacts with noise levels 6 dB and 10 dB in which the output SNR increased around 9 dB or more. For baseline drifts, this method performs very well for noisy signal with noise levels 2 dB, 6 dB, and 10 dB.