Combination of PNN network and DTW method for identification of reserved words, used in aviation during radio negotiation

Vitaliy J. Larin; Evgenii E. Fedorov

doi:10.3103/S0735272714080044

Authors

Vitaliy J. Larin National Aviation University, Ukraine https://orcid.org/0000-0002-5042-2426
Evgenii E. Fedorov Donetsk Academy of Motor Transport, Ukraine

DOI:

https://doi.org/10.3103/S0735272714080044

Keywords:

probabilistic neural network, mel-frequency cepstral coefficients, discrete wavelet transform, probability, identification, aviation, air traffic management

Abstract

It is proposed a method of Bayes estimation of the results of cross-segmentation of observed and reference vectorial sequences in different spaces of signs, which are applied for speech encoding in a systems of key words recognition. It is considered a case of identification of reserved words used in aviation during radio negotiations of an air traffic controller with a pilot or an operator with control system of pilotless vehicle on a basis of combination of PNN with DWT method. Carried out experimental researches using a vocabulary of 45 words with test sample of 4500 words allow to conclude probability of reserved words recognition is achieved greater than 0.95.

References

FEDOROV, E.E. Methodology of Development of Multi-Agent System of Speech Control. Donetsk: Knowledge, 2011 [in Russian] 356 p.
ATAL, B.S. Automated recognition of a speaker by voice. TIIER, v.64, n.4, p.48-66, 1976.
HAYKIN, S. Neural Networks: A Comprehensive Foundation, 2nd ed. New Jersey: Prentice-Hall, 1999, 1104 p.
RABINER, L.R.; JANG, B.H. Fundamentals of Speech Recognition. New Jersey: Prentice Hall PTR, Englewood Cliffs, 1993, 507 p.
BOURLARD, H.; MORGAN, N. Connectionist Speech Recognition: A Hybrid Approach. Boston: Kluwer Academic Publishers, 1994, 312 p.
RABINER, L.R.; SCHAFER, R.W. Digital Processing of Speech Signals. New Jersey: Prentice Hall, Englewood Cliffs, 1978, 495 p.
KASURIYA, SAWIT; WUTIWIWATCHAI, CHAI; ACHARIYAKULPORN, VARIN; TANPRASERT, CHULARAT. Comparative study of continuous hidden Markov models (CHMM) and artificial neural network (ANN) on speaker identification system. Int. J. Unc. Fuzz. Knowl. Based Syst., v.9, n.6, p.673, 2001, DOI: http://dx.doi.org/10.1142/S0218488501001149.
VINTSYUK, T.K. Analysis, Recognition and Interpret of Speech Signals. Kyiv: Naukova Dumka, 1987 [in Russian] 261 p.
BEZVESILNA, O.M.; LARIN, V.Y.; CHICHIKALO, N.I.; ET.ALL. Converting Equipment Devices. Technological Measurements and Devices. Zhytomyr: ZhDTU, 2011 [in Ukrainian] 542 p.