V.R. Krasheninnikov and A.I. Armer (Russia)
Speech command, speech signal, pronounciations,variation, simulation, autoregressive image model,speech recognition
the corresponding reference SC. However, often one and the same SC pronounced by the same announcer at different time has significantly different characteristics (signal length, level and form). These differences make the SC comparison procedure difficult, which mistakes and discrepancies reduce probability of correct SC identification. Therefore at present development of ways of SC variability description and simulation is being conducted. Nowadays there exist a number of speech formation models ([1], [2], [3]) but satisfactory methods for analytical description of SC variability are still not found and statistical description requires processing of large amount of data. In this paper an approach to SC variations modeling pronounced by an announcer using Habibi's autoregressive image model. Nowdays the mass communication and service systems focused on the speech management, and also reliable the person and the computer dialogue systems find broad application. Designing of these systems is based on the development in the speech recognition field. The speech command variations simulation at different pronounciations offered here may serve for speech recognition systems designing and testing simplification.
Important Links:
Go Back
