Auscultation is still the main procedure by the physician in determining the health condition of a person’s lungs. Auscultation heavily depends on the physician’s skill and experience. Electronics auscultation using computer assistance is used to identify abnormalities in lung sounds for reducing the subjectivity. One of the signal processing methods that is often used to determine the lung sounds is wavelet decomposition method. This study aimed to compare several methods of lung sound classification using wavelet analysis. Some methods combined wavelet decomposition techniques and features extraction to obtain a method that produces the highest accuracy with the fewest number of features. The results showed that the DWT order 7 with DB2 mother wavelet and 46 features produce the highest accuracy of 97.98%. This method was tested on five classes of lung sound data.

wavelet transform, lung sound, feature extraction, multilayer perceptron.