Improvement of ECG Signal Quality Using Normalized Unbiased Adaptive Filter

This paper presents a normalized unbiased adaptive noise reduction system for improving the quality of electrocardiogram signals. The proposed system consists of four stages: a two-stage moving average filter for baseline wander removal, an IIR notch filter for power line interference suppression, Gaussian white noise injection as a reference signal, and a normalized adaptive filter based on the gradient descent algorithm for eliminating high-frequency random noise. Unlike conventional methods that require detailed signal feature analysis such as QRS complex detection, the proposed model operates without the need to identify these features. The adaptive process is designed to minimize the instantaneous error between the estimated signal power and the noise-free signal, while the filter coefficients are maintained in normalized form. Performance evaluation was conducted using the MIT-BIH database at various noise levels. Results demonstrate that the proposed model exhibits superior performance compared to the LMS filter and is suitable for clinical applications.

Keywords:

Adaptive filter , Notch filter , Gradient descent