Classification of Obstructive and Restrictive Respiratory Diseases Based on Cough Sounds

Respiratory diseases are a leading cause of mortality, with obstructive and restrictive conditions posing significant diagnostic challenges due to overlapping symptoms. This paper explores machine learning and deep learning techniques, such as K-nearest neighbor (KNN), Convolutional Neural Networks (CNN), long short-term memory (LSTM), support vector machine (SVM), CNN-LSTM, and ensemble methods, to differentiate between obstructive and restrictive respiratory diseases based on cough sounds. Mel frequency cepstral coefficient features are utilized across all models. Results show the ensemble model achieves the highest accuracy of 98.63%, outperforming KNN (95.89%), CNN (93.15%), LSTM (80.95%), SVM (92.68%), and CNN-LSTM (93.15%).