A simple real-time QRS detection algorithm utilizing curve-length concept with combined adaptive threshold for electrocardiogram signal classification

QRS detection is a standard procedure in electrocardiogram (ECG) signal classification and analysis. Although there is a large number of methods published, some featuring high accuracy, the problem remains open. This is especially true with respect to high accuracy QRS detection in noisy ECGs such as long-term Holter monitoring during normal daily activity. In this paper a robust real-time QRS detector for noisy applications is proposed. It exploits a modified curve-length concept with combined adaptive threshold derived by basic mean, standard deviation and average peak-to-peak interval. The method was tested using the MIT-BIH arrhythmia database with an observed detection accuracy of 99.70%, sensitivity of 99.86%, positive prediction of 99.84%, and an average failed detection of 0.30%. The proposed approach compares favourably with published results for other QRS detectors, and proves superior to those having constant and manually entered threshold parameters.