Title screen

[References: 11 tit.]

The problem of the cardiac signal extraction from a low-amplitude high-frequency component invisible on conventionalelectrocardiography (ECG) is considered. For this purpose, the original ECG model, which includes a highamplitudecomponent, a low-amplitude component, and random noise, is used. We have proposed the method toextract the orthogonal high-amplitude and low-amplitude components from the recorded cardiac pulse sequence inthe form of the first two components of the matrix singular decomposition composed of a set of cardiac pulses. Incontrast to Simson’s method, the proposed method does not assume cardiac signal averaging over a large (aboutseveral hundred cardiac pulses) sequence as well as the averaged signal filtering. The low-amplitude component isextracted from a short (about 10 s) ECG recording. We have used a mathematical model to examine the stability of theproposed method to the correlation of the high-amplitude and low-amplitude components of the cardiac signal as wellas to the influence of random noise. It is shown that the high-amplitude component is steadily extracted in the entirerange of the change in noise level and correlation of the regular components of the cardiac signal. At the same time,the low-amplitude component is significantly distorted when the pair correlation coefficient with the high-amplitudecomponent is greater than 0.2 or noise level is greater than 0.4.