Background: Electrocardiogram signal (ECG) is a graphical representation of the heart activity. Processing and analysis of these morphological changes can result in visual diagnosing some cardiac diseases. However, various types of noises and disturbances in ECG influence the visual recognition and feature extraction from it. The aim of this research is to eliminate different noises from ECG and to enhance its quality.

Materials and Methods: In this study, an adaptive Kalman filter is developed by using Bayesian model. Considering simplification and Gaussian distribution for measurement noise, complicated mathematical equations were converted to simple relations and therefore implementation was simplified.

Results: In this paper, by designing an adaptive Kalman filter, the signal to noise ratio (SNR) has increased to 21.46dB. Adaptive Kalman filter based on Beyesian framework could model dynamic variations of ECG signal by estimating covariance matrix for measurement noise.

Conclusion: In despite of Kalman filters that use parametric functions to model ECG signal, the adaptive Kalman filter introduced in this paper uses real ECG records for modeling. Parametric functions which could model dynamic variations of ECG need a lot of analytical functions and this decreases the time of filtering process but the adaptive Kalman filter proposed in this research has a high speed and could be used in real time applications.