Spectrum analysis of respiratory sound: application to smokers and non-smokers.

Previous studies have indicated that respiratory sound signals may contain information useful in the detection of lung diseases. In this study, measurement and recordings of respiratory sound signal segments were obtained in normal subjects (non-smokers) and smokers in both inspiration and expiration phases. By using the autoregressive (AR) method, it is possible to produce power spectra of respiratory sound signals in inspiration and expiration phases for smokers and non-smokers of each group. The selection of the AR model order of the respiratory sound signals is achieved using Akaike criterion. The AR model order of 9 is required for completely described respiration sound signal segments in inspiration and expiration phases for both groups. The power spectra in the smoker group show larger distinct peaks at lower frequencies as well as more harmonics in both inspiration and expiration phases compared to the power spectra of the non-smoker group. Another diagnostic indicator was derived from the relative position of poles of the AR model of respiratory sound signals. In all smokers it was found that the first, third and fourth poles were closer to a unit circle than those in non-smokers (P < 0.01). It seems that the use of these indicators may be useful as early diagnostic tool for lung diseases.