Machine Learning-Based Respiratory Disease Classification Using Lung Sounds

Asthma, bronchitis, pneumonia, normal and chronic obstructive pulmonary disease (COPD) all call for timely and dependable detection if diagnosis and treatment are to be effective. We present a respiratory sound classification system that combines digital signal processing with machine learning. Lung sound recordings drawn from publicly available respiratory sound datasetsfirst undergo noise reduction, normalization, and audio format conversion. From every processed recording we thenextract twenty Mel-Frequency Cepstral Coefficients (MFCCs) that capture its acoustic character, and these values are passed to a Random Forest classifier that assigns the recording to a respiratory disease category. A Flaskbased web application wraps the trained model and offers audio upload, preprocessing, disease prediction, confidence-score display, and waveform and spectrogram visualization. When evaluated with standard classification measures, the Random Forest classifier reached an accuracy of accuracy: 99.21% ,which supports its use for automated early screening of respiratory sounds. Keywords—Respiratory Sound Classification,Lung Sound Analysis, Mel-Frequency Cepstral Coefficients (MFCC), Random Forest, Digital Signal Processing, Machine Learning, Respiratory Disease Detection.