A. S. Krasichkov, E.V. Frolova, A.N. Akulov

Synthesis and enhancement of R-wave detection algorithms is one of the key points in biological signals monitoring and online analysis systems design. The recorded ECG is often corrupt by noise of various origin. Significant improvement in the algorithm can be achieved by considering the individual parameters of patient's ECG that are determined either during preliminary investigation or during the first minutes of multi-lead ECG recording (learning mode). The paper suggests an R-wave detection algorithm in multi-lead ECG synthesized in accordance with the statistical hypothesis testing theory where the decision is made when the statistics exceeds a certain thre-shold. The decision statistics is based on the weighted analysis of cross-correlations between the current input at each of ECG leads and the corresponding PQRS-complex samples, also with compensation of cross-correlation between the samples and base line drift. The weights are assigned according to the energy characteristics and noise level in each lead. Once the decision statistics exceeds a fixed threshold, a local maxima search procedure is performed. It has been found that those samples with highest energy perform the best. The algorithm efficiency has been checked at three-lead ECG (V3, V5 and V6) with artificially added simulated myographical noise and base line drift. The suggested algorithm has demonstrated high efficiency of QRS-complex detection at high noise levels compared to conventional algorithms.