摘要 |
FIELD: technological processes. ^ SUBSTANCE: invention is related to restoration of speech, more precisely, to method and device for restoration of speech in system of distributed speech recognition. Method and device are suggested for speech restoration in system of distributed speech recognition. Omitted coefficients MFCC are restored and used for generation of speech, in particular, partial restoration of omitted MFCC coefficients is achieved by application of dependence of omitted MFCC coefficients on transmitted period P of the main tone, and also on transmitted MFCC coefficients. Amplitudes of harmonics are then produced from transmitted and restored coefficients of MFCC, and speech is restored with application of these amplitudes of harmonics. ^ EFFECT: improved restoration of speech by means of increase of accuracy of harmonics amplitudes transformation. ^ 8 cl, 4 dwg |