I.V.Vovk, A.Ya.Kal'uzhny
On potential efficiency of spectral-correlation methods of classification of breath sounds

Acoustic bulletin, Vol. 2 1, (1999) p.11-21
The optimum algorithm of the respiratory noise classification on base of spectral-correlation criterions has been considered and the parameters of its efficiency have been determined. The numerical analysis of such parameters for idealized model of human respiratory system has been performed. There has been investigated the comparative diagnostic value of model parameters including the conditions of the respiratory noise generation and the acoustic characteristics of medium of propagation. The advantages of spectral-correlation methods of noise classification in comparison with the spectral ones have been shown. The robustness of qualitative parameters of system of classification to the discrepancy in knowledge of parameters of respiratory system and errors in determination of spatial coordinates of acoustic sensors have been investigated.
KEY WORDS:
classification of breath sounds, the spectral-correlation methods
TEXT LANGUAGE: Russian