Japanese vowel devoicing has been studied from the point of view of both computer speech processing and speech science. Present knowledge of devoicing is based upon data of Tokyo (standard) Japanese. However, since frequency of devoicing occurs less often in the Kinki dialect than in the Tokyo dialect, information on various dialects in which devoicing occurs less frequently is needed in order to have fully applicable speech processing techniques. Recently, we have examined vowel-devoicing phenomena in the Tokyo- and Kinki dialects using traditional methods of speech analysis, as well as fiber-optic viewing of the larynx. In this paper; we present results of our recent studies as well as preliminary results of a newly-started speech data analysis of newspaper scripts.
展开▼