Preprocessing PPG and ECG Signals to Estimate Blood Pressure Based on Noninvasive Wearable Device

Abstract

Accurate systolic and diastolic blood pressure measurement is still an open problem in biomedical engineering. This paper introduced a method how to preprocess the PPG and ECG signals to get a series of parameters to estimate blood pressure. It developed a threshold detector to extract characteristic points of PPG and ECG waves and calculated several useful features such as pulse wave velocity. The paper also introduced a BP-feature based model by using those features to estimate SBP and DBP. The accuracy of detection algorithm reached 98.2% by comparing with SFM database. The estimation results were validated by a large-scale dataset we acquired by strict experiment procedures. The average deviation error in estimating SBP and DBP was 0.25 and 2.24 mmHg respectively. The standard deviation between the measured and predicted blood pressure was 8.92 mmHg for systolic pressure and 8.13mmHg for diastolic pressure, which met the standard of AAMI. The results indicated that the BP-feature based model has a reliable estimation of blood pressure.