Validation of Algorithm with Noise Tolerance Methods to Detect R-wave

Abstract

Long-term monitoring of electrocardiogram (ECG) is one of the basic measurement in healthcare and provides decisive information regarding cardiovascular system status. In all ECG applications, the R-wave detection is important. However, it is difficult to detect R-wave automatically because signals obtained in daily life frequently include noise from various sources. Daily life monitoring ECG signal is particularly measured over cloth during walking and sleeping states by using non-invasive sensor, so that it usually presents higher noise level. To improve detection accuracy under noisy condition, we developed algorithm which has some noise tolerance techniques that analyze characteristics of R-wave. The proposed algorithm was evaluated by using the records of the MIT-BIH Polysomnographic database and the data from nonintrusive vital sign monitoring system previously developed in our laboratory. Algorithm reliability was assessed by detection error rate (De), sensitivity (Se) and positive predictivity (P+). The result shows average De of 2.62%, average Se of 98.29% and average P+ of 99.03%. We suggest that our R-wave detection method is useful in the presence of noisy signals.