| Abstract [eng] |
In this work of diphthongs spectral analysis is briefly discussed the past and the present of sounds synthesis. Also discussed about humans speech organs and main language generation principles. Available formant variation analysis, and formant methods. It discusses what could be speech signal generation model. The study describes process of diphthongs spectral analysis. Words where selected from dictionaries, their acquisition and digital processing was accomplished with program PRATT. This speech signal processing system gives user these opportunities: enter speech signal to computer, see signal graphic in a screen, edit signals, hear entered signal or wanted fragment, see spectrogram, formantive rated frequency, the basic tone of the period graph and main spectrum segment, to filter signal, write down signal characteristics to file. In experiment participated six people: three women and three men. They phrase 81 word with 7 diphthongs. Speech was recorded in silent inveronment, but not in a professional environment. Labeling of phonemes was accomplished with PRAAT program. Phonemes were cut manually taking into account the inherent length of each phoneme. After designation for each word where created Text Grid file, which holds in itself information about each target volume threshold and duration. Using software Matlab is gotten spectrograms of diphthoning words used by people participating in experiment, those spectrograms are analyzed. In conclusion is submitted what diphthoning formants changes depends on which accentual intonation is used in word’s accent. |
