ECG Signal Reconstruction Using Interacting Multiple Model

Abstract

The Electrocardiography (ECG) is the method of reading the electrical movement of the heart over time by keeping electrodes on a body of a person. ECG signals are frequently studied to diagnose possible diseases in human body. It has become common for any medical evaluation and has been used for a long time. In recent days, computer-assisted ECG analysis is playing an important role in helping doctors in the study and treatment of heart abnormalities. an ECG can be used to measure the rate and rhythm of heartbeats, the size and position of the heart chambers, the presence of any damage to the heart’s muscle cells or conduction system, the effects of cardiac drugs, and the function of implanted pacemakers. ECG is used to check the health of the patient’s heart when other diseases or conditions are present. These include high blood pressure, high cholesterol, cigarette smoking, diabetes, and a family history of early heart disease. This makes the ECG signal the most enduring tool for the cardiologist. Many ECG signal modeling techniques have been proposed for compression and classification.
Polynomial Modeling of an ECG waveform is one of the oldest methods. But the main drawback of polynomial models is that we have to use the same type of polynomial for each point. However, fiducial points vary, depending on the person and the person’s cardiac health. That problem can be overcome by using different polynomial models to model the same ECG waveform which can be done by using the interacting multiple models (IMM) framework. The multi-mode property of this framework allows us to switch different modes of operation using first-order Markovian transition probabilities. Dynamical systems can be modeled with few possible modes of operation instead of single by using the IMM algorithm. The IMM algorithms use Kalman filters which are run in parallel. These individual filters are initialized by using results obtained from previous steps. Finally, the overall estimate can be given by mixing the estimates of individual filters. The IMM algorithm can be described using the interaction, filtering, and combination stages. This IMM algorithm provides us with the polynomial coefficients for each and every point in the ECG waveform. Similarly, we can extend this framework by replacing Kalman filter with an extended Kalman filter