Classification of Lung Sound Signals using Machine Learning

Although the easiest pre-screening approach for lung disease diagnosis is using a stethoscope to analyse the lung sounds, it is not employed as a conclusive diagnostic test because it is greatly dependent on the physician's competence and expertise. This research presents a machine learning-based autonomous lung disease detection system in the hope of making lung sound diagnosis more conclusive. After signal decomposition with Empirical Mode Decomposition (EMD), the Ensemble Rus-Boost classifier is employed in this method to classify lung sounds. Chronic Obstructive Pulmonary Disease (COPD) vs non-COPD, COPD vs healthy, COPD vs URTI and COPD vs Pneuomonia are the four binary classifiers analysed in this paper.