SPEECH SYNTHESIZER

Japanese Patent JP3302874

B2

PROBLEM TO BE SOLVED: To lessen the deterioration in sound quality and to form natural synthesized speeches approximate to human speeches by executing preselection in accordance with the vocal sound information and accent information of unit speech data, forming unit speech candidate sets and selecting the optimum unit speech data from these sets.
SOLUTION: The character strings inputted to a synthesis unit forming section 2 are segmented to the unit speeches which are the units for forming the synthesized speeches. These unit speeches are sent to a preselection section 3. In the preselection section 3, the preselection of the unit speech data from a unit speech data base 4 is executed with each of the unit speeches in accordance with the vocal sound information and accent information of pronunciation symbol strings. The unit speech data collected by the preselection section 3 from the unit speech data base 4 are sent as the unit speech candidate sets to a unit speech data selection section 5. In this unit speech data selection section 5, the optimum unit speech data is selected and is sent to a unit speech extraction section 6. As a result, the synthesized speeches having high naturality are formed.

JPH09160582	VOICE SYNTHESIZER
JPH01118200	VOICE SYNTHESIS SYSTEM
JP5336026	Memory addressing method for systems with variable speed data playback function

Keiko Nagano

JP4837496A

July 15, 2002

February 09, 1996

Click for automatic bibliography generation  

NEC

G10L13/02; G10L13/06; G10L13/07; (IPC1-7): G10L13/06; G10L13/08

JP63237099A
JP573092A

Naoki Kyomoto (2 outside)

Previous Patent: CLEANER AND CLEANING METHOD

Next Patent: REACTIVE POWER COMPENSATOR FOR PROTECTING SYSTEM INTERCONNECTION