Identification of Pulmonary Disorders by Using Different Spectral Analysis Methods

This study presents detection of pulmonary disorders using different spectral analysis methods such as fast Fourier transform, autoregressive and the autoregressive moving average. Power spectral densities of the sounds were estimated through these methods. Feature vectors were constructed by extracting statistical features from the PSDs. Created feature vectors were used as inputs into the artificial neural networks. Then performances of spectral analysis methods were compared according to classification accuracies, sensitivities and specificities. In this aspect, the study is a comparative study of different spectral analysis methods.