As there is no fundamental frequency in whisper, abstraction of other acoustic characteristics from whispered Mandarin becomes more prominent in speech analysis and synthesis. In this paper, the differences between whispered and voiced vowels in Chinese are statistically analyzed. At first, the formant of whisper is discussed. Then, Gain in whisper is described. Next, tone recognition for whispered Mandarin is descanted. Experimental results with six mono-vowels /a/, /o/, /e/, /i/, /u/, and /ü/ indicate the following: (1) the first and second formant frequencies of whispered vowels show shifts toward higher frequencies, but that trend is only strongly evident for F1, (2) the formant bandwidths are expanded and are nearly constant throughout the all formants, (3) the gain of whispered vowel is much lower than that of voiced vowel, and (4) the envelope of whispered spectrum might carry some tonal information.
[1]
Kazuya Takeda,et al.
Analysis and recognition of whispered speech
,
2005,
Speech Commun..
[2]
H. Traunmüller,et al.
Dept. for Speech, Music and Hearing Quarterly Progress and Status Report a Comparative Study of the Male and Female Whispered and Phonated Versions of the Long Vowels of Swedish
,
2022
.
[3]
K. Kallail,et al.
Formant-frequency differences between isolated whispered and phonated vowel samples produced by adult female subjects.
,
1984,
Journal of speech and hearing research.
[4]
E. Paulus,et al.
Speech Signal Processing
,
1997,
The Electrical Engineering Handbook - Six Volume Set.
[5]
Hideki Kasuya,et al.
Acoustic nature of the whisper
,
1999,
EUROSPEECH.