To provide a speech synthesizer which is used to output high quality speech with a natural pitch pattern.
A speech element/prosody control section 23 moves the approximate shape of a basic pitch pattern, that is produced by a prosody production section 22, along a frequency axis direction in parallel while holding the shape, searches speech elements for every synthesis unit to compute an evaluation value and obtains an overall evaluation value of the entire deformed pitch pattern. Then, the deformed pitch pattern having a best overall evaluation value is set as a pitch pattern that is to be used during speech synthesis and the speech elements that are selected during the above operation is set as the speech elements to be used during the speech synthesis. Thus, synthesized speech without great degradation in speech quality is produced in each synthesis unit for the best deformed pitch pattern that is finally obtained. Moreover, the deformed pitch pattern is set by conducting frequency deformation while holding the approximate shape of the basic pitch pattern without considering a deformation possible range. Thus, high quality speech is outputted with a natural pitch patter.
COPYRIGHT: (C)2004,JPO
Kimura Osamu
Tomoichi Morio
JP2002196780A | ||||
JP2000194390A | ||||
JP2000066694A |
Hiroshi Yamazaki
Next Patent: METHOD AND DEVICE FOR PROCESSING MOVEMENT ADAPTIVE TYPE SIGNAL