Modeling of respiratory crackles

Crackles, extracted from lung sounds recorded from pathological subjects, are parameterized using Prony modeling of order two. The scatter plots of Prony parameters show that there is a correlation between the type of pathology and the lung volume of crackle occurrence and the Prony frequencies. These distinctive distributions of crackle populations of each disease indicate that Prony modeling generates crackle parameters which are promising features to be used in the classification of lung sounds along with the more conventional feature set of AR or ARMA model parameters.