[Clustering analysis of phonocardiagrams]

Through the development of the ART-2 neural network, we had constructed the 2-dimensional ART neural network appropriate for clustering the time-varying spectrums. After that, according to the character of phonocardiagram(PCG), we chose the scalogram of PCG to be the input feature of the network. By adjusting parameters the network was enabled to cluster the input PCGs to two clusters. The result demonstrates that the system we designed can pick out, from among 32 continuous PCGs, two PCGs which have the tendency toward abnormality. This result may be applicable to basic physiological studies and clinical diagnosis.