Blood Pressure Prediction Based on Pattern Classification

摘要

With the development of wearable medical care, some researchers pay more attention to wearable blood pressure estimation. In general, the photoplethysmography(PPG) signal is used to extract the featuresfor regression analysis. However, although many regression methods (e.g., Linear regression, ANN, SVR, etc.) can be applied in this field, how to improve the accuracy of blood pressure prediction is a big problem. In this article, based on pattern classification, a method of blood pressure estimation is firstly proposed. Compared with the previous regression analysis, the different ranges of blood pressure are divided into different classes. Then the signals are classified into the corresponding category according to the extracted features. Blood pressure value could be obtained by finding the interval median. Experiments have shown that the idea of classification can be able to achieve the higheraccuracy of blood pressure prediction.