Classification of blood volume pulse signals using an artificial neural network Bayesian classifier

The assessment of the condition of the cardiovascular system through morphological analysis of the blood volume pulse (BVP) was extended to include the use of an artificial neural network (ANN). The dicrotic notch feature of the BVP was used to define 4 classes and an ANN Bayesian classifier was used to make the assignments. Training and testing data were obtained from a clinical study. Resting and exercise BVP data were collected from 15 individuals and these signals were normalized prior to being input to the ANN. Several ANN configurations were evaluated and percent correct classification (pcc) rates >90% were obtained from the 5, 7 and 5-5 hidden layer configurations. These results are compared with those from K nearest neighbor and Parzen window classifiers.