Support Vector Machine for Abnormal Pulse Classification

Radial pulse signals have been utilized in ancient culture health diagnosis due to its simple, non invasive approach. Characteristics of a newly identified abnormal pulse in the subjects suffering from gastritis and arthritis are along with commonly visible healthy pulse patterns in this work A binary classifier to segregate such abnormal pulse healthy pulse patterns is modeled using linear, quadratic as well as support vector machine based algorithms. Frequency domain features derived from power spectral density of the pulse signal are ranked to achieve dimensionality reduction. It has been found that the support vector machine with linear kernel classifies the abnormal pulse signals with highest success rate of 99.2% utilizing only two ranked frequency domain features. General Terms Linear Classifier, Quadratic Classifier, Support Vector Machine, Kernel Function, Dimensionality Reduction