V.V.Krizhanovskii, V.V.Krizhanovskii (jr.) Adaptive respiratory signal processing under limited a priori information |
Acoustic bulletin, Vol. 11 ¹ 1, (2008) |
| A problem of synthesis of an adaptive algorithm for detection and classification of respiratory sounds in difficult interference conditions is considered within the limits of a statistical approach. To estimate the interference characteristics, a training data sample was used. We have analyzed the variants of rational forming of the training sample of a background interference on the basis of the records containing the intervals with pauses in respiration. The sufficient statistics of the adaptive respiratory sounds detector is determined and its structure is analyzed. It is shown that respiratory sound preprocessing by the adaptive filter of synthesized detector allows using of traditional methods for estimating of the informative frequency band and some other parameters important for signal classification. A procedure is offered for the adaptive hierarchical classification of respiratory sounds. The results of experimental verification are presented for the efficiency of adaptive detecting of typical narrow- and broadband respiratory sounds. The efficiency of use of the adaptive classification procedure for subdividing of respiratory signals into classes of the sounds, describing the norm and pathology of organs of respiration, is estimated experimentally. The schemes for hierarchical classifying of basic and complementary respiratory sounds are offered. |
|
KEY WORDS: respiratory sounds, signal detection, interferences, adaptive algorithm, hierarchical classification |
| TEXT LANGUAGE: Russian |