Artificial neural networks for acoustic lung signals classification

Abstract

A classification of acoustic lung signals for the respiratory disease diagnosis problem is studied in the present work. Models based on artificial neural networks, using Mel Frequency Cepstral Coefficients for training are employed in this task. Results show that neural networks are comparable, and in some cases better, with other classification techniques as Gaussian Mixture Models, that work on the same database.