Filatova N.N., Khaneev D.M. Respiratory noise recognition algorithm based on the neural class models

Abstract: The article describes an algorithm for detection of respiratory noise, based on the idea of growing pyramidal network adapted to operate with fuzzy descriptions of objects in the learning samples set and enriched with the linguistic interpreter for the processing results. The article contains a general functional diagram and detailed description of the individual stages of work. To describe the symptom space and the interpretation of results authors used the theory of fuzzy sets. The functioning of the algorithm is carried out in two modes: training and recognition. Neural classes models, contained in the constructed network, are interpreted in fuzzy statements, which are then used in the learning mode and provide a set of production rules for the algorithm of fuzzy logical inference. The given algorithm has a software realization, the article presents the results of testing its software implementation.

Keywords: