| 摘要 |
<P>PROBLEM TO BE SOLVED: To diversify features of a speaker for synthetic voice generated by voice synthesis. <P>SOLUTION: Voice parameter registration processing takes a voice waveform inputted in a performance period of an object musical piece, as the object to be processed. Voice waveform data SV based on the voice waveform is divided into syllables included in vocalization contents in association with performance periods of respective output sounds constituting a melody line of the object musical piece, to generate syllable waveforms (S140), and a voice parameter PM is derived from each syllable waveform (S150). Furthermore, periods when the same key continues (namely, sections of the same key) in the object musical piece are specified respectively to specify the key (tonality) in the sections of the same key, and a voice quality preliminarily prescribed as an emotion imagined from the specified key is specified as metadata (S160). Thereafter, voice parameters PM and metadata are stored in a data storage server in association with each other per corresponding syllable (S170). <P>COPYRIGHT: (C)2013,JPO&INPIT |
