NOISE REMOVAL FROM ECG SIGNAL USING SSRLS ALGORITHM

Abstract

Electrocardiogram (ECG) is the basic tool for diagnosis of heart condition of a patient which reflects the electrical activity of the heart. The ECG signal may be affected by different types of disturbances/noises e.g. Power Line Interference (PLI), baseline wander, motion artifacts etc. Out of all the types of noises, PLI is the main contributor to ECG signal distortion. Removing such disturbing signals from ECG is the key to its accurate analysis leading to diagnosis of potential disease(s). In this work, an adaptive approach to PLI tracking and elimination from ECG signals using State Space Recursive Least Squares (SSRLS) algorithm has been presented. This thesis work has been divided into three parts depending upon the type of PLI noise being tracked by SSRLS algorithm. First part deals with the case where frequency of PLI noise is known in advance. Second part deals with the case where PLI noise frequency is unknown but has a fixed value. Last part of thesis deals with the case where PLI noise is drifting in any arbitrary manner. All the parts of the thesis have been simulated using SSRLS algorithm and results have been compared with other adaptive techniques. Finally, the performance of SSRLS algorithm has been tested by using practical ECG signal taken from Armed Forces Institute of Cardiology (AFIC) database and appreciable results have been achieved. The two parameters of SSRLS algorithm namely forgetting factor λ and step-size parameter η plays an important role in the performance of SSRLS algorithm in all the considered cases.