DENOISING OF ECG SIGNAL USING ADAPTIVE INDEPENDENT COMPONENT ANALYSIS

It has been observed that denoising of ECG is done in stationary condition i.e. when the patient under scanner is lying on bed or in rest condition. This is very common case and the noise added in this condition can be removed by many different methodologies like FFT, wavelet, ICA etc. However, when the ECG is monitored in dynamic condition e.g., on tread mill testing, the noise inserted due to motion artifact is of entirely of different nature as compared to that of the static condition. In the presented work, the focus is on denoising of ECG signal corrupted by motion artifact conditions. During the acquisition of ECG signal it gets corrupted due to different types of artifacts and interferences that may hide important diagnostic information. Independent Component Analysis (ICA) is a blind source separation technique that can be used for the removal of such noises and artifacts. In this paper, different ICA schemes such as JADE algorithm and Fast ICA are discussed and applied for ECG denoising. The database used is MIT-BIH database. ECG recordings are examined by a physician who visually checks features of the signal and estimates the most important parameters of the signal. Using this expertise the physician judges the status of a patient. Therefore the recognition and analysis of the ECG signals is a very important task. The standard parameters of the ECG waveform are the P wave, the QRS complex and the T wave. But most of the information lies around the R peak. Additionally a small U wave (with an uncertain origin) is occasionally present. Related Works In this paper a combination of Extended Kalman Filter (EKF) and a dynamic model of a synthetic electrocardiogram (ECG) for ECG denoising is proposed. Experimental results show that the proposed algorithm is very efficient for the extraction of the ECG signals from noisy data measurements (Ouali et al., 2013). In this paper an efficient filtering procedure based on the singular Value Decomposition (SVD) has been proposed. SVD, a high resolution spectrum estimation tools, is used to decompose the ECG data matrix into orthogonal subspaces. Due to the energy-preserving orthogonal transformation in the SVD, these subspaces correspond to the signal and noise components contained in the ECG data. Projection of the data onto the desired subspace eliminates the noise and the unwanted signal components (Ouali and Chafaa, 2013). Noise always degrades the quality of ECG signal. ECG noise removal is complicated due to time varying nature of ECG signal. As the ECG signal is used for the primary diagnosis and analysis of heart diseases, a good quality of ECG signal is necessary. A survey of various types of noises corrupting ECG signal and