Home – Home – Jornals – Advanced Search

It is shown that for the p-order autoregressive process, in the case of using a prediction optimized for minimum root-mean-square error, and the known autoregressive coefficients, the predictability time can exceed the correlation time by a factor of p for the best combination of autoregressive coefficients.

The problem of processing measurements containing dynamic observation interferences is solved. The obtained method is invariant to piecewise continuous interferences of deterministic structures with unknown parameters. The method does not require extending the state space and speeds up solving the estimation problem.

Selective effects of the centroid transformation method for structural analysis of monochrome graphic images are investigated. The principle of centroid transformation with the use of functionals of the type of moments for constructing space-structure parameters is described. Two processing schemes are considered: a centroid value accumulation scheme and a cluster scheme. It is shown that when the image is affected by noise, even a high-intensity noise, the scheme of detecting structure elements and borders via clustering the values of parameters is quite efficient. Analytical expressions defining the coordinates of centers and the sizes of clusters, and also the values of rms deviations of the parameter values as a function of rms deviations of the value of uncorrelated additive image noise are found.

Nonparametric pattern recognition algorithms based on a randomized optimization approach are proposed. The idea of the approach is to validate the random character of fuzziness factors of kernel functions and choose parameters of their distribution law by optimizing nonparametric decision rules. Properties of the developed qualifiers are investigated. Results of their comparing with conventional nonparametric pattern recognition algorithms are analyzed.

Problems of synthesizing neural-network models under incomplete experimental data are described. The accuracy of the models is heavily dependent on their complexity. The proposed method allows self-organizing neural-network models of near-optimal complexity. Examples of synthesizing models of flame interferometry and growth of industrial production are presented.

Angular selectivity of volume phase transmission gratings formed in a liquid crystal composite for s- and p-polarized light is investigated in the absence of external electric field.

Numerical modeling is performed in order to extract the fluence and energy for the case of dielectric SiO₂ layers implanted by Si⁺ ions from spectroscopic ellipsometry measurements. To this end, one-, two-, and three-layer models are tested. It is found that a two- or three-layer model is sufficient for determining the implantation parameters and dielectric layer thickness with acceptable accuracy. Further complication of the model does not improve the accuracy, but greatly complicates the search procedure for minimization of objective function.

Single-frequency traveling-wave Nd:YAG and Yb:YAG lasers combining the functions of intracavity second harmonic generation and birefringent filter on one nonlinear crystal are discussed. The lasers were developed for spectroscopy and metrology applications. Output characteristics and spectral tuning ranges of the lasers are presented.

A new technique of low-coherent interferometry is discussed. In the method, the optical field reflected from the layered medium is directed into a longitudinal share scanning interferometer for autocorrelation analysis. The object under analysis is outside the interferometer and the reference beam is not used. Theory, results of numerical simulation, and experimental data are presented.

The possibility of visualizing the fine structure of optical inhomogeneities in gas and condensed fluids by Hilbert optics methods is discussed. Application of optical Hilbert filtering for investigating shock waves and dynamic perturbations induced by the waves in gas is described.