350 rub
Journal Science Intensive Technologies №9 for 2010 г.
Article in number:
An algorithm of optimal structural approximation of experimental multidimensional signals
Authors:
A.G. Dmitriev
Abstract:
By now there are many articles dedicated to the analysis of structural signals being processed in the majority of cases in two stages: singling out «one type» fragments (a stage of segmentation) and giving then a general description of the signals presented. Application of the existing methods is not effective enough, when the research is carried out under conditions of considerable dimensions, limited experimental observations and the unknown number of «one type» fragments (intervals of partitioning). To solve this problem it is suggested in the article to use a method of introducing a criterion of quality for the decisions being taken. The decision to be found is one that gives the extreme value to the criterion chosen. Using the advantage of the introduced criterion over the usual criterion of quadratic misalignment - an extreme character of its dependence on the number of intervals of partitioning, an algorithm of determining the number of intervals of partitioning has been worked out. To determine limits of intervals of partitioning one uses method of dynamic programming that makes it possible to get the global minimum of the criterion. The choice of the final variant of the approximating function is made by the use of statistics according to the Fisher distribution law. Statistics is used to check the adequacy of the models at the complementary intervals of partitioning making it possible to get true estimates of the of the parameters of the local models of the approximating function.
Pages: 31-35
References
- Моттль В.В., Мучник И.Б. Скрытые марковские модели в структурном анализе сигналов. М.: Физматлит, 1999. 352 с.
- Дмитриев А.Г. Методы анализа и описания сложных кривых. ЧВИИРЭ. 2009. 77 с. - Деп. в ЦВНИ МО РФ 12.05.2009, № В7039
- Дмитриев А.Г., Дорофеюк А.А. Методы кусочной аппроксимации многомерных кривых // Автоматика и телемеханика. 1984. № 12. С. 101 - 108.
- Беллман Р., Дрейфус С. Прикладные задачи динамического программирования. М.: Наука. 1969. 457 c.