A semiautomatic labeling system for known speech sample is proposed to construct a fine database for speech research. An acoustically compact phonetic unit called demiphoneme is introduced, and a word (or sentence) is represented by a network using demiphonemes which cover the acoustic variation contained in the utterances of the word (or sentence). The system performance is experimentally evaluated by the labeling of 36 city names uttered by 10 male speakers.
|Denshi Gijutsu Sogo Kenkyusho Iho/Bulletin of the Electrotechnical Laboratory
|Published - 1986
ASJC Scopus subject areas