No abstract

New samples of protective garment for performing construction work at height require numerous tests in conditions close to real conditions of extreme vital activity. The article presents some results of shockproof garment element studies and a description of a patented prototype. The tests were carried out on a model which geometric dimensions were convenient for manufacturing it in a limited batch. In addition, the used laboratory equipment (for example, a unique power pendulum), blanks made of a titanium-nickel alloy with a shape memory effect also imposed their limitations. The problem of the adequacy of the obtained experimental results transfer to mass-produced products was solved using tools of the classical similarity theory. Scale factor management influence in the affine modeling of the shockproof element, studied on the basis of the equiatomic titanium-nickel alloy with the shape memory effect, allowed us to assume, with a sufficient degree of reliability, the technical possibility of extrapolating the results of experimental studies to full-scale objects for the formation of the initial data of the mathematical model of shockproof garment dynamics elastoplastic deformation (while observing the similarity of the features of external loading).

No abstract

Abstract. We propose a method for controlling Gaussian mixture splitting in HMM states during the training of acoustic models. The method is based on introducing special criteria of mixture quality in every state. These criteria are calculated over a separate part of the speech database. We back up states before splitting and revert to saved copies of the states whose criteria values have decreased, which makes it possible to optimize the number of Gaussians in the GMMs of the states and to prevent overfitting. The models obtained by such training demonstrate improved recognition rate with a significantly smaller number of Gaussians per state.

No abstract