350 rub
Journal Biomedical Radioelectronics №4 for 2012 г.
Article in number:
Complicated objects classification method based on analysis of structural functions of slow waves
Authors:
I.I. Volkov, S.G. Emelyanov, S.A. Filist
Abstract:
There is complicated systems classification method based on analysis of structural functions of slow waves. The method is concluded in a priory choice of slow waves - continuum, determination of slow waves structural functions, included in the continuum, and using these structural functions as informative attributes of the decision-making block, built on neural network technology. The method allows to make as control of complicated objects - dynamics as their classification.
The decision-making block of objects classification was designed to realization of the method. Work of the block is based on analysis of structural functions, received as a result of slow waves - extraction from nonstationary signal. Slow waves cluster, based on samples of given slow waves continuum of current and etalon classes of system state. System dynamics control is done by determining the Euclidean distances of each set of structural functions samples and following neural network analysis of the informative features - resulting space.
There is an algorithm of counting structural functions of nonstationary signal - slow waves at the article. Essence of the algorithm is that the k-th sample of structural function is evaluated by iterative process, realized in the inner cycle. Moreover amount of iterations doesn-t depend from the number of slow wave, on which structural function, allowing to receive rectangular matrix with the structural functions - samples of chosen slow waves continuum of the system, is built.
Firstly samples of the slow waves continuum are compared with the same samples of similar continuum of the l-th class - etalon object to forming the l-th class - characteristic space. Then the distance between present state of object and its - control state is calculated, what allows to classify objects in two and more classes and to control state dynamics of the same object.
Special action is used to control of trajectory of object in the space of informative characteristics. Its - efficiency is measured by definition of reaction - changing the distance between present state of object and its - control state before and after the action. In depend of the result they correct direction of control impact, what allows to realize stable operation of complicated object - trajectory in the state space of slow waves.
Pages: 6-12
References
- Белобров А.П., Кузьмин А.А., Филист С.А. Многомерная частотная селекция в задачах анализа медленных волн // Биомедицинская радиоэлектроника. 2010. № 2. С.4 - 10.
- Белобров А.П. Нейронная сеть для анализа структурных функций медленных волн // Сб. материалов XIV междунар. научн.-техн. конф. «Медико-экологические информационные технологии-2010». Курск: ЮЗГУ. 2011. С.31 - 42.
- Прохоров С.А., Графкин В.В. Структурно-спектральный анализ случайных процессов // СНЦ РАН. 2010.