Yacko

The article presents estimation results of automatic speech recognition by the system based on the method of hidden Markov models. The research was carried out in the following conditions: in a laboratory in the absence of acoustic noises, under the influence of acoustic noises of different types. The presented results give objective quantitative estimation of such characteristics as speaker-independence and noise-immunity of speech recognition systems with little vocabulary, based on Markov models; recommendations on improving speech recognition characteristics are formulated. The authors propose further development of the techniques developed for estimation of influence of flight overloads on speech characteristics and its automatic recognition.