Automatic isolation and analysis of nasals and nasalized vowels in continuous speech
| Autor: | H. B. Ritea, R. A. Gillmann |
|---|---|
| Rok vydání: | 1974 |
| Předmět: |
Acoustics and Ultrasonics
Computer science Acoustics Speech recognition Linear prediction respiratory system behavioral disciplines and activities Formant Linear prediction model Arts and Humanities (miscellaneous) Vowel behavior and behavior mechanisms otorhinolaryngologic diseases Spectral analysis Formant bandwidth Nasal vowel psychological phenomena and processes |
| Zdroj: | The Journal of the Acoustical Society of America. 55:S21-S21 |
| ISSN: | 0001-4966 |
| DOI: | 10.1121/1.1919597 |
| Popis: | Nasals and nasalized vowels comprise approximately 14% of the phonemes used in English Speech. An algorithm is described for locating and identifying nasals and preceding nasalized vowels. Spectral analysis is performed on a computer using a linear prediction model containing both poles and zeros, to locate the formants and antiformants of these phones. The effectiveness of this approach is compared to that of an all pole (LPC) model, using formant bandwidth data to locate the antiformant. The identification of nasalized vowels is based on the hypothesis that the formants of an unnasalized vowel, when nasalized, tend to move away from the antiformant of a succeeding nasal. Experimentation has indicated that (1) automatic techniques can be employed to isolate nasals and preceding nasalized vowels in continuous speech; (2) linear prediction can be used to effectively analyze the spectra of these phones; and (3) speaker‐dependent tables of formants and antiformants extend these results to multiple speaker en... |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|