O.V. Rodionov - Dr.Sc. (Eng.), Professor, Head of Department, Voronezh State Technical University
S.A. Borisovsky - Ph.D. (Eng.), Research Scientist, Research Center of the Federal State Unitary Enterprise «18 Central Research Institute» Ministries of Defence of the Russian Federation (Kursk)
Ya Zar Do - Post-graduate Student, Southwest State University

Currently complex-structured signal processing is widely used in biology and medicine. To singular analysis of complex-structured signal is a representation of the signal in the form of additive segments and each segment of the relevant signs intended to build classifying models (neural networks, fuzzy decision rules, etc.). Advantage of this approach is that the informative signs is structured into blocks that correspond to physical or physiological properties of the object of study, allowing for their classification not only methods that reveal the statistical properties of the learning samples, but also the expertise of the experts. Signal decomposition method is suggested for the additive components in proportion to the main components of the covariance matrix of trajectory of the matrix of the original signal. It is established that the application of the proposed method allows not only to group the additive components, but also to filter the signal by removing part of the additive components, corresponding to certain private numbers. The singular analysis of us made èlektrokardiosignal (ex) with a duration of 2:0, selected from the database of the European society of Cardiology (ESC ST-T database). Offered in four interactive environment of additive components: trend, fast recurrent processes periodic processes are slow and chaotic. To perform classification grouped additive components in the online environment uses Visual analysis of psevdotraektornyh matrices of singular decomposition of the trajectory of the matrix of the original signal. Interactive analysis of additive components generate one additive components. The informative signs for training neural network classifiers generated by building models, such as polynomial or spectral of additive components. Singular analysis applied to classification of complex-structured of electrophysiological signals in decision support systems for diagnostics of cardio-vascular diseases.