Automated System Design for Classification of Chronic Lung Viruses using Non-Linear Dynamic System Features and K-Nearest Neighbour

Breath sounds provide substantial information about the state of a respiratory system. These signals are frequently influenced by noises from heart and muscles which complicate the research process. Chronic Lung Virus (CLV) is a very common lung disease that necessitates an accurate diagnosis at the outset to receive adequate care. A smart framework for diagnosing breath sounds-based viruses is proposed in this study. Breath signals will first be preprocessed by Interval Dependent Denoising (IDD), which divides a signal into intervals by different variance shift points, and then filtered by the Savitzky-Golay (SG) filter that eliminates high-preference muscle noise and components. Next, five Non-Linear Dynamic System (NLDS) features are extracted. Initially, 5-fold cross-validation is used to train and validate the K-Nearest Neighbor (KNN) classifier. The suggested solution is experimentally assessed on several different categories of a locally collected dataset from Pakistan that includes 1454 breathing sounds from healthy and chronic lung virus (flu virus, rhinovirus) affected subjects. KNN with medium kernel obtained 99.4% accuracy, 99.99% sensitivity, 97.82% specificity and 99% precision.