سامانه همایش ها

English

صفحه اصلی / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

A Novel AR-Based Kalman Filtering Framework for ECG Enhancement

نویسندگان :

Hamed Danandeh Hesar¹

1- Sahand University of Technology

کلمات کلیدی :

ECG denoising،autoregressive model،Kalman filter

چکیده :

Model-based Bayesian methods, while effective for preserving ECG morphology, often demand extensive preprocessing (e.g., R-peak detection, phase assignment, offline parameter optimization), hindering their real-world use. Our prior adaptive Kalman filter bank reduced preprocessing to R-peak detection but introduced computational complexity and iterative parameter adjustments via expectation maximization. To address these limitations, we introduce a novel approach: an autoregressive (AR) state model embedded within a two-stage Kalman filter framework for ECG denoising and parameter estimation. Critically, this method eliminates the need for any preprocessing, including R-peak detection. It achieves this by adaptively estimating model parameters and dynamically adjusting state and measurement noise covariances based on intrinsic noise characteristics. Extensive evaluations on benchmark datasets (MIT-BIH Normal Sinus Rhythm, Arrhythmia; Creighton University Ventricular Tachyarrhythmia; Sudden Cardiac Death Holter) reveal that our proposed AR-based Kalman smoother (AR-KS) surpasses state-of-the-art model-based adaptive Kalman filter banks, particularly for signal-to-noise ratios (SNRs) above 0 dB.