Heart beat detection method with estimation of regular intervals between ECG and blood pressure

Heart beat is a fundamental cardiac activity which is easily found with various measurements. In the worst case, unexpected noise or contaminated signals can affect ECG signal in the practical conditions, then, we hardly find the heart beat for a while. To deal with these problems, PhysioNet/CinC suggests a challenge for developing robust heart beat detection method with multimodal data. In this paper, we proposed two models which process EEG and a pair of ECG and BP signals to compensate the limitation of QRS detection. At first, we extract candidates of heart beat using QRS detector. As supplementation, we estimated regular interval between ECG and BP signal. When we miss heart beats on ECG, we can estimate the locations of heart beat using detected characteristic BP signal and regular interval. At last, we decomposed EEG signal which contains ECG signal as cardiac artefact. Adaptive filtering model in cascade was proposed for this purpose. We utilized the model to identify the locations where the cardiac artifacts are occurred on EEG signal. Then, voting module generated the suggested locations of candidate heart beats. Our proposed method achieved the overall score with 82.01% at phase III.