Radial arterial pulse is an important physiological signal that has been applied in Traditional Chinese Medicine (TCM) for thousands of years.From ancient times, pulse has been recognized as an empirical science and plays a decisive influence on the TCM diagnosis.However its objective and lack visible database, which blocks the development of TCM.In Recent years, many pulse systems based on various kinds of sensors have been introduced to collect the computerized pulse waveforms.Meanwhile, pulse diagnosis using statistical learning theory is attracting more and more attention.This paper mainly presents the pulse feature extraction algorithm for removing the redundant and irrelevant information.Though many researches on pulse feature have been published, most of them emphasize on a certain aspect and hardly utilize the experience in TCM.We propose an integrated framework of pulse features and introduce the corresponding extraction algorithms.The experiments show that the features are extracted accurately and they performance well in disease diagnosis.