A Fuzzy Human Model for Blood Pressure Estimation

The paper describes a blood pressure prediction model. The model predicts blood pressure of the subject based on trend of the blood pressure, body weight and number of steps. To predict it, we make autoregressive (AR) model, liner prediction model, body weight based prediction model and steps based prediction model. These models are boosted by fuzzy logic. The fuzzy degrees are calculated from mean absolute prediction error, correlation coefficient and variation amount for the learning data. In our experiment, we collected blood pressure, body weight and number of steps of 453 subjects from WellnessLINK which is an internet life-log service. Our proposed model predicted their blood pressures. The mean correlation coefficient between the predicted values and measurement systolic blood pressures was 0.895.