Estimation of heart rate variability from heart rate measured with wristwatch-type heart rate monitor

Abstract

HF and LF/HF, which are parameters of the frequency domain of the electrocardiogram R-R interval time series, are general index of the autonomic nervous activity balance. Most wristwatch-type heart-rate monitors measure heart rate from sphygmograph and interpolate missing data. These are the causes of reducing the calculation accuracy of LF and HF. The purpose of this study is to propose a method for estimating LF and HF calculated from the ECG R-R interval time series from the heart-rate time series at 1-s intervals measured by a wristwatch-type heart-rate monitor. Six hundred and seventy data-set with a length of 120 s measured simultaneously with an electrocardiogram and a wristwatch-type heart-rate monitor during standing cooking operation, up-right standing, watching slide show in standing posture, and resting supine position were used. Seventy percent dataset were used for modeling, and remaining 30% were used for evaluating model. As a result, the coefficient of determination of LF and HF were 0.441 and 0.467, respectively in the one-variable polynomial regression model. The coefficient of determination of LF and HF were 0.557 and 0.644, respectively in the multivariable regression model, based on Random Forest and time domain indices.