发明名称 JAPANESE SPEECH SYNTHESIZING METHOD
摘要 PROBLEM TO BE SOLVED: To determine a sound-continuing time length of an individual phoneme in a short time and prevent a statistic error from locally enlarging, by previously performing a category classification for all music element symbols, and determining the sound continuing time length for each input music element in an input music element symbol row, based on a category of the input music element included in the input music element symbol row. SOLUTION: For example, when a phoneme symbol row corresponding to an input text 'Good morning, Taro' is 'taro'saN%ohayo', a phoneme to be noted is 'N', and a phoneme next to the noticed phoneme is a pause '%' (S31). Therefore, a first search table is searched, based on two phonemes before and after the noticed phoneme (S314). An average sound-continuing time length of the noticed phoneme and a second search table are derived from the first search table (S3141). The mord number in an exhalation stage including the noticed phoneme, and the average sound-continuing time length of the noticed phoneme by the discrimination of the end of a sentence, whether the noticed phoneme lies within three morde are adjusted (S3142-S3144).
申请公布号 JP2000214876(A) 申请公布日期 2000.08.04
申请号 JP19990016002 申请日期 1999.01.25
申请人 SANYO ELECTRIC CO LTD 发明人 NISHIDA HIDEJI;HIRAI HIROYUKI;HASHIMOTO MAKOTO
分类号 G10L13/08;G10L13/06;(IPC1-7):G10L13/06 主分类号 G10L13/08
代理机构 代理人
主权项
地址