| Authors |
Volchikhin Vladimir Ivanovich, doctor of technical sciences, professor, president of Penza State University (40 Krasnaya street, Penza, Russia), president@pnzgu.ru
Ivanov Alexander Ivanovich, doctor of technical sciences, associate professor, head of biometric and neuronal nets technology laboratory, Penza Scientific Research Electrotechnical Institute (9 Sovetskaya street, Penza, Russia), ivan@pniei.penza.ru
Malygina Elena Aleksandrovna, candidate of technical sciences, research assistant, interdisciplinary laboratory testing of biometric devices and technologies, Penza State University (40 Krasnaya street, Penza, Russia), mal890@yandex.ru
Serikova Yuliya Igorevna, postgraduate student, Penza State University (40 Krasnaya street, Penza, Russia), julia-ska@yandex.ru
|
| Abstract |
Background. The aim of this work is to describe the procedure of symmetrization of the correlation of multidimensional biometric data through the computation of the full correlation matrix and elements with the same value.
Materials and methods. The symmetric correlation matrix are not amenable to classical regularization by Tikhonov, but they have their own structural stability. It turns out that the symmetrization of the correlation is not that other, as a certain method of structural regularization for the solution of multidimensional problems of linear algebra.
Result. It is shown that the transition to the calculation of the symmetric correlation functionals of order n equivalent to the increased volume of biometric data in n(n − 2)/ 2 times compared to the computation of conventional correlation coefficients. Increasing the dimensionality of the correlation functionals leads to a monotonous decrease of the error of their calculations arising due to lack of source data.
Conclusions. Correlation functionals is not that other, as a multidimensional convolution in space of the input States are available for monitoring statistics. In practice widely used only the simplest two-dimensional correlation functionals, since the use of the correlation functionals of higher order is poorly studied. A barrier that prevented earlier, start researching possibilities vysokorazvityh correlation functionals were expectations for a decline of the accuracy of their calculations. Such concerns are valid only in relation to asymmetric correlation functionals.
|
