Pattern Classification for Doppler Ultrasonic Wrist Pulse Signals

Wrist pulse signal contains important information about the health status of a person and it has been used in Traditional Chinese Medicine for a long time. In this work, digitalized wrist pulse signals from patients with different diseases as well as healthy persons are collected by a Doppler ultrasonic device. Two methods, namely, the wavelet method and the auto regressive prediction error (ARPE) method, are proposed to analyze the pulse signals and distinguish patients from healthy persons. Distinctive features are first extracted from the pulse signals and then the support vector machine (SVM) is used for classification. The applicability of the methods is investigated using wrist pulse signals collected from 50 healthy persons and 74 patients. The results illustrate a great promise of the proposed methods for computerized pulse signal analysis.