Computer programms implement the adaptive algorithms for ECG signal real-time filtering

In this version of the computer programs, the data type has been changed from “integer” to “longint” for processing longer signals. Multilevel noise estimation and, respectively, adaptive switching of a larger number of filter sets are used. The filter parameters are adjusted by numerical simulation for a very wide range of variance of the additive Gaussian noise from its absence and very high input SNRs to their negative values. An increase in the number of possible parameter values that can be adaptively switched during processing improves the filters dynamic and statistical properties, and does not significantly decrease the processing speed. The algorithms parameters are given in the “filters.txt” file. Optionally, the number of parameters can be reduced by setting the same filter parameters for the next sets. For the parameters adjustment an optimization algorithm was not used. Therefore, the parameters only close to optimal have been selected. A typical ECG cycle is used as a model signal for numerical simulation and evaluation of the filter efficiency. As examples, the parameters of the proposed filtering algorithms are adjusted for the signals from the NSTB and PTB Physionet databases at the sampling rates of 360 Hz and 1000 Hz. The advantages of the proposed algorithms for non-stationary noise suppression in ECG are their high efficiency and low processing delay, allowing high-speed performances in real time mode