ABSTRACT
Chronic Obstructive Pulmonary Disease (COPD) is one of the largest causes of death in the world. This chronic lung disease is caused by lung damage that can no longer be cured. This disease is very closely related to cigarette smoke that is inhaled either by smokers or by those around them who inhale the smoke for a prolonged time as passive smokers. There are many symptoms for COPD, such as shortness of breath especially after exercise, wheezing, cyanosis, cough, and cough with phlegm. Our data consist of 20 patients with COPD and 15 healthy control patients. Cough sound was recorded for 10 minutes using a clip-on microphone. Then, we separated every cough into one-second segmentations and extracted the Mel-Frequency Cepstral Coefficient (MFCC) feature vectors. We trained the data using two machine-learning algorithm classification models: support vector machine and multi-layer perceptron. We achieved 0.94 as the best result of both sensitivity and specificity.
