V.V.Krizhanovskii, V.V.Krizhanovskii (jr.), V.A.Chernets Experimental estimates of the efficiency of the respiratory sounds dichotomy under partial a priori indeterminacy |
Acoustic bulletin, Vol. 10 ¹ 1, (2007) |
| The paper deals with the problem of the respiratory sounds dichotomy to the norm and pathology according to their power spectral density in the conditions of limited a priori information. The main principles of the system approach are formulated and classification attributes, efficient for solving this problem are determined. The classification algorithms coordinated with preset attributes are offered and their informative characteristics are determined. The adaptive Bayesian procedure of the classification algorithm adjustment in the corresponding informative space of attributes is determined. The procedure is proposed for allocating the informative frequency band of the respiratory sounds according to pauses. Its characteristics, important for solving the dichotomy problem, are determined. The efficiency of the suggested classification procedures has been experimentally tested and the obtained results have been analyzed. The opportunity of increasing the dichotomy efficiency was shown, when simultaneously using the results of operation of different algorithms. |
|
KEY WORDS: respiratory sounds, diagnostics of diseases, an informative characteristic, probabilistic algorithms, adaptive classification, rank classification |
| TEXT LANGUAGE: Russian |