Electrocardiogram denoising using Wavelet decomposition and EMD domain filtering

During the data acquisition and transmission of Electrocardiogram (ECG) signal in the clinical environment, the signal gets corrupted by various types of noises e.g. additive white Gaussian Noise (AWGN), electrode contact noise, electromyography (EMG) noise, etc. For accurate diagnosis, denoising of ECG signal is necessary as it affects the diagnostic features. The proposed denoising approach combined the two methods, wavelet and empirical mode decomposition (EMD) for removal of AWGN. In this method, noisy ECG signal is primarily decomposed using wavelet decomposition resulting in approximation sequence and detail sequence. The soft-thresholding method is then used to denoise the detail sequence. This gives the denoised detail sequence which, in conjunction with approximation signal gives partially denoised signal. The partially denoised signal have observed to have denoised QRS complexes but the remaining part of this signal still have noises. The EMD decomposes this signal into several IMFs. The denoised QRS complex is then detected and preserved from the first three IMFs. The noisy segments of the remaining IMFs (other than first three IMFs) is then detected with the help of instantaneous half period (IHP) and filtered using soft-thresholding technique. The proposed Wavelet & EMD approach is evaluated using the records from MIT-BIH Arrhythmia Database.