Acoustic Characteristics of Japanese Vowels

The purpose of this study was to build a large database on Japanese vowels and to clarify their acoustic characteristics. Recordings were made of 256 males and 252 females aged from 6 to 76 years old, for both Kanto (Tokyo) and Kansai (Osaka) dialects, producing the vowels /i, e, a, o, u/ in isolation, in /h-Vda/ syllables and in /b-V-ta/ syllables within and without a carrier sentence. The five vowels uttered with different notes were also recorded. A total of 154,570 vowel tokens were collected and phoneme labels were given to each of them. Analysis of the formant data showed that there was little difference between the two dialects. Formant frequencies were the highest for children's vowels and those of boys and girls were comparable in distribution. Formant frequencies of men were the lowest. The mean F0 of men dropped at the age of 12 to 13 due to the voice breaking. Formant frequencies shifted up when vowels were uttered at higher musical notes. The rate of the shift due to F0 change, however, was smaller than that seen among age groups. Dynamic features of formants varied with phonemic environment but were not affected by mora structure and talkers age.