15th International Congress of Phonetic Sciences (ICPhS-15)

Barcelona, Spain
August 3-9, 2003

Fundamental Frequency Analysis of Longitudinal Recording in a Japanese Infant Speech Database

Shigeaki Amano, Tomohiro Nakatani, Tadahisa Kondo

NTT Corporation, Japan

To investigate the acoustic characteristics of spoken language development, the utterances of five Japanese infants and their parents were recorded from the infants' birth until the age of 60 months. An infant-speech database is now being developed from these longitudinal recordings. New methods of fundamental frequency (F0) extraction and voiced/unvoiced segment detection were proposed in order to obtain the F0 of infant speech, which has a wide frequency range at a high background noise level. A preliminary analysis of the F0 revealed that with infant speech there is a slight tendency for the F0 to decrease and that Japanese parents use a higher F0 with infants under 20 months of age. This use of a higher F0 is one of the characteristics of infant-directed speech as reported for other languages. However, other characteristics of infant-directed speech were not observed, such as exaggerated intonation contours and a slower speaking rate. It is suggested that the higher F0 is a dominant feature of Japanese infant-directed speech and that infant-directed speech has different characteristics across languages.

Full Paper

Bibliographic reference.  Amano, Shigeaki / Nakatani, Tomohiro / Kondo, Tadahisa (2003): "Fundamental frequency analysis of longitudinal recording in a Japanese infant speech database", In ICPhS-15, 1983-1986.