Abnormality Detection in Lung Sounds Using Feature Augmentation

Assessing the lung sounds reveals important information about the lungs and the existence or severity of possible underlying respiratory related conditions. This paper presents a research-work-in-progress exploring a hybrid signal processing and machine learning-based approach for effectively analyzing the lung sounds. The objective of this study is to achieve high accuracy in detecting adventitious sounds for pulmonary diseases non-invasively. To this end, we are augmenting the feature space and introduce a feature-based model. For the analysis, we are utilizing the ICBHI dataset, which comprises lung sounds collected from 126 patients. The dataset includes various abnormalities, such as wheezing and crackling sounds, providing us with valuable data to train and evaluate the proposed model.