Robust time series processing for Heart Rate Variability analysis in daily life

Heart Rate Variability (HRV) analysis can be heavily affected by slow trends and outliers in inter-beat interval data, especially when recorded during exercise. In this paper, a trend and outlier removal method is presented. Inter-beat interval data are considered as time series and decomposed into trend, outlier and fluctuation components using an iterated algorithm. Trend components reflect heart rate variation due to activity or exercise and are estimated based on Empirical Mode Decomposition (EMD). Outlier components reflect noise and other abrupt interference and are estimated using non-parameter method. The fluctuation components are obtained by removing trend and outlier components from original heart rate data, and thus have pure heart rate variation caused by autonomous nervous system function and are used for HRV analysis. This iterative method can handle heart rate data contaminated by large percentage of outliers and slow trends simultaneously, which is usually the case during exercise.