V.A. Aladinskiy – Ph.D.(Eng.), Professor, Scientific Secretary, 

LTD «STC» (St. Petersburg)

E-mail: awa2810@yandex.ru

S.V. Kuzminskiy – Ph.D.(Eng.), Scientific and Pedagogical Worker,  Cherepovets Higher Military Engineering School of Radio Electronics

E-mail: cvviur6@mil.ru

D.N. Chubaty – Ph.D.(Eng.), Associate Professor, 

Cherepovets Higher Military Engineering School of Radio Electronics E-mail: cvviur6@mil.ru

A method for selecting informative features of recognition of low-speed speech coding (LSSC) protocols implemented in vocoders, which consists in accepting a binary digital information stream Y during the time interval T, forming a normalized autocorrelation function A based on Y, making a decision about the presence of a block structure in the information flow Y according to regular extrema of the auto-correlation function A at equal intervals Δτ, divide the information flow Y into information blocks with a volume of Nb bits each, at intervals between the extremes of the autocorrelation function A, sequentially assign information blocks k = 1, 2, …, K to the information blocks, starting from the first information block, form a rectangular information matrix Y, L = Nб; whose rows are, in succession, information blocks placed under each other in accordance with their serial numbers k = 1, 2, …, K. A sequence of columns y1, y2, …, yz, …, yL is formed, the current number is z = 1, 2, …, L of the column yz is determined by the column number of the rectangular information matrix, the mathematical expectation mz is calculated for each of the yz columns, the vector of the measured mathematical expectation values m = {m1, m2, …, mz, …, mL} is formed on the basis of obtained set of values of mathematical expectation mz their proper placement in accordance with serial numbers z, form a set of vectors of measured values of mathematical expectation M = {m (0), m(1), …, m(l), …, m(L-1)} by shifting the values of the vector m by the value l = 0, 1, 2, …, L-1, form J reference vectors of the values of the mathematical expectation, j = 1, 2, …, J, the known LSSC protocols for J training samples, w = 1, 2, …, W; containing W digital information flows, each lth vector of measured values of mathematical expectation m (l) is compared with each j-th, j = 1, 2, …, J, a reference vector of values of mathematical expectation of known LSSC protocols, and the probability of correct recognition of the LSSC protocol Pj(l), according to the maximum probability of correct recognition of the LSSC protocol Pj(l) decide in favor of the j-th LSSC protocol.

The method according to claim 1, characterized in that any known alternative set of recognition features of the LSSC protocols is used, corresponding to the format x = {x1, x2, …, xz, …, xL} and having distinctive properties, is formed according to the training alternative reference vector of values {Yj,w}W, one by one, for each pair of LSSC protocols, the degree of information content of each of the alternative sets of signs is determined based on the calculation of the divergence values, the decision to use one of the sets of signs is made by maksimal value divergence between pairs of reference vectors.

The method according to claim 1, characterized in that in a situation with the number of classes J > 2 and with equal values of Nb for all classes under consideration, the average divergence between all classes is calculated, the decision to use one of the sets of signs is made according to maximum mean divergence.

1. Aladinskii V.A., Kuzminskii S.V. Analiz tsifrovykh potokov na vykhodakh vokoderov, primenyaemykh na zarubezhnykh liniyakh radiosvyazi diapazona vysokikh chastot. Uspekhi sovremennoi radioelektroniki. 2015. № 7. S. 71−74. (in Russian)
2. Alekseev A.A., Kirillov A.B. Tekhnicheskii analiz signalov i raspoznavanie radioizluchenii. SPb.: VAS. 1998. 368 s. (in Russian)
3. Pat. RF na izobretenie № 2667462 ot 19.09.2018. Sposob raspoznavaniya protokolov nizkoskorostnogo kodirovaniya rechi. Aladinskii V.A., Veshchunin E.A., Kuzminskii S.V., Smirnov P.L. Federalnaya sluzhba po intellektualnoi sobstvennosti, byull. № 26. 19.09.2018. (in Russian)
4. Tu Dzh., Gonsales R. Printsipy raspoznavaniya obrazov: Per. s angl. M.: Mir. 1978. 411 s. (in Russian)
5. Timofeev D.I., Tavalinskii D.A., Chubatyi D.N. Analiz parametrov nizkoskorostnykh koderov rechi v usloviyakh strukturnoi i parametricheskoi neopredelennosti. Naukoemkie tekhnologii. 2011. № 8. S. 4−9. (in Russian)
6. Aladinskii V.A., Kuzminskii S.V. Metod formirovaniya priznakov raspoznavaniya protokolov nizkoskorostnogo kodirovaniya rechi. Naukoemkie tekhnologii. 2015. № 12. S. 20−25. (in Russian)
7. Pat. RF na izobretenie № 2610285 S1 ot 08.02.2017. Sposob raspoznavaniya protokolov nizkoskorostnogo kodirovaniya. Aladinskii V.A., Kuzminskii S.V., Smirnov P.L., Chubatyi D.N. Federalnaya sluzhba po intellektualnoi sobstvennosti, byull. № 4. 08.02.2017. (in Russian)
8. Zagoruiko N.G. Metody raspoznavaniya i ikh primenenie. M.: Sov. radio. 1972. 208 s. (in Russian)