Robust measurement of English vowel formants in stressed monophthongs: A comparison of single- and multi-point methods
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study investigates the most robust method for measuring vowel formant frequencies in stressed monophthongs. Five approaches were compared using the Aix-MARSEC database: measurement at the midpoint (Mid), at the one-third point (OneThird), at three equally spaced points (Three), across all points between one-third and two-thirds of the interval (MidAll), and across all points throughout the vowel interval (Full). Measurements of F1Bark, F2Bark, and Euclidean distances of individual vowels from the centroid of each vowel were analyzed using standard deviations, correlation tests, and linear mixed-effects regression models. Results indicate that the Full method—averaging formants at 10-ms time steps across the entire vowel interval—consistently yields the most stable and reliable measurements. This robustness is evident once vowels following /j/, /w/, or /ɹ/, and those preceding /l/, are excluded. In this study, outliers are included or removed, and only vowels exceeding 40 ms in duration are retained. The findings challenge the traditional reliance on single-point measurements and demonstrate that multi-point analysis, particularly the Full method, offers a more reliable representation of vowel formant frequencies. These results highlight the importance of comprehensive temporal sampling in acoustic phonetics.
Abstract
This study investigates the most robust method for measuring vowel formant frequencies in stressed monophthongs. Five approaches were compared using the Aix-MARSEC database: measurement at the midpoint (Mid), at the one-third point (OneThird), at three equally spaced points (Three), across all points between one-third and two-thirds of the interval (MidAll), and across all points throughout the vowel interval (Full). Measurements of F1Bark, F2Bark, and Euclidean distances of individual vowels from the centroid of each vowel were analyzed using standard deviations, correlation tests, and linear mixed-effects regression models. Results indicate that the Full method—averaging formants at 10-ms time steps across the entire vowel interval—consistently yields the most stable and reliable measurements. This robustness is evident once vowels following /j/, /w/, or /ɹ/, and those preceding /l/, are excluded. In this study, outliers are included or removed, and only vowels exceeding 40 ms in duration are retained. The findings challenge the traditional reliance on single-point measurements and demonstrate that multi-point analysis, particularly the Full method, offers a more reliable representation of vowel formant frequencies. These results highlight the importance of comprehensive temporal sampling in acoustic phonetics.