Traditional approaches are based on the selection of features that have the highest correlation with BP. In this paper, we propose a novel method for non-invasive blood pressure (BP) estimation which is based on pulse transition characterization using Electrocardiography (ECG) and Photoplethysmography (PPG) signals. The study was assessed on a large surgical dataset of 2152 subjects. Compared with previous work, we show that pulse characterization is necessary for more accurate BP estimation.