A New Measure of Pulse Rate Variability and Detection of Atrial Fibrillation Based on Improved Time Synchronous Averaging

Figure 3

Steps of preprocessing. During period segmentation, the original pulse wave series were segmented into periods by threshold method. During signal quality evaluation, the segment obtained by period segmentation were divided into normal and abnormal segments by a logistic regression model. The abnormal segments were eliminated, and the range of normal segment was expanded by 50% on both sides to include the information of adjacent periods.