Search academic texts

Information × Registration Number 0211U000932, 0109U002907 , R & D reports Title Development of automated computer system for digital phonogram phonoscopic examination. popup.stage_title Head Shelepov Vladislav Yurievich, Registration Date 03-03-2011 Organization Institute of Artificial Intelligence Problems popup.description2 Object of research: digital audio of acoustic and none acoustic character. Subject of research: methods and algorithms of digital audio signal processing, which include lowering effect of noise in the signal and voice activity detection, and methods and algorithms for speaker identification. Goal of research: to develop of methods for automated processing of speech and speaker identification, that is based on existing methods of digital signal processing and pattern recognition. The results that are achieved: improved technology for signal preprocessing on purpose of voice activity detection, noise reduction and analysis of phonograms availability to the identification study, methods of speaker identification features formation that are stable to GSM-compression, improved technology of melodic contour obtaining and definition of intonation type syntagmas in linguistic analysis framework, software product in the form of information technology of automated of speech signals processing in forensic expertises of digital phonograms framework. Scientific novelty: a voice activity detection method, that takes into account the spectral characteristics of speech sounds, which prevents the possibility of adopting short-term high-amplitude noise for a speech or a speech signal for the low-amplitude noise, the spectral subtraction method is modified, this method takes into account the acoustic features of speech sounds, what allows to determine the boundaries of speech in signal and hold it to the segmentation of voiced, unvoiced areas, noisy unvoiced obstruent sounds, the technique of determining intonational type syntagma, taking into account mikroprosodical effects, this method is based on the technique of obtaining intonational characteristics in the form of time series, that are used in the speech synthesis; set of identification features is modified by using psychoacoustic scales to represent the local values of the signal power, thus improving the efficiency of identification by comparison with a standard method. Scope of application: developed speech signal processing methods and algorithms in addition to forensic expertises systems may be used in speech synthesis system, determine the emotional state of the speaker, and biometric identification. Product Description popup.authors Єрмоленко Тетяна Володимирівна Панфілов Сергій Станіславович Панфілова Марина Олексіївна Шелєпов Владислав Юрійович popup.nrat_date 2020-04-02 Close