Time-varying wiener filtering based on short-time fourier transform

The paper considers filtering noisy pulse signals using Wiener filtering technique applied in the Short Time Fourier Transform domain. It is assumed that there exists some a priori estimation of the signal spectrum within some time interval for each moment, such estimation may be available when the signal consists of some transients repeated in quasi-periodical manner, or when there exists some concurrent much less noisy process that can be used to estimate the momentary spectrum of the filtered signal. Efficiency of the proposed technique is demonstrated on filtering noisy magnetocardiographic signals based on concurrent electrocardiograms. Its superiority over stationary Wiener filtering is shown.