Analysis and Automatic Classification of Breath Sounds

Analysis of breath sounds (auscultation) is an important part of the diagnosis of pulmonary diseases. An automatic breath-sounds classification scheme is suggested. Types of normal and abnormal breath sounds are classified, with the goal of providing the physician with a diagnostic assist device. The classification is performed in two levels. The first level is based on the linear prediction coefficients, and the second level on energy envelope features. Seven types of breath sounds were classified. Out of 105 trials, the system exhibited only five errors. The system can be implemented on a microprocessor and thus can be offered as a low-cost clinical instrument.