Automatic detection and classification of acoustic breathing cycles

This paper focuses on respiratory phase detection and classification without the help of the airflow measurements. Instead of using the airflow measurements to identify breathing phases, the proposed work depends on advanced digital signal processing techniques to process the acoustic signal of respiration that was collected using a microphone placed in front of the subject´s nose. The recorded signal is processed using the voiced-unvoiced algorithm to differentiate between the voiced period and the unvoiced (silence) period. The desired features are extracted from each voiced phase. Finally, support vector machine is used to distinguish between the inspiration and the expiration phases according to the extracted features. The signals were recorded from a number of subjects who do not have a history of pulmonary diseases. The proposed method has achieved an accuracy of 95% when tested on the subjects.