Method and device for preparing and using diphones for multilingual text-to-speech generating

• US 5,897,617 A
• Filed: 08/14/1996
• Issued: 04/27/1999
• Est. Priority Date: 08/14/1995
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Collier, Rene P. G.
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): SAX, ROBERT L
• Application Number:
  US08/696,431
• Time in Patent Office: 986 Days
• Field of Search: 704/260, 704/258
• US Class Current: 704/260
• CPC Class Codes: G10L 13/06 Elementary speech units used in speech synthesisers; Concatenation rules
• CPC Class Codes: G10L 15/063 Training

Abstract

Diphones are prepared for text-to-speech converting by selectively pronouncing a set of selected diphones and processing each such diphone for persistent storage. Finally, each processed diphone is stored in an individually addressable manner. In particular, amongst such set as spoken by a single person, on a basis of homophony each diphone is assigned to one or more diverse languages. Sharing of selective diphones amongst more than one language diminishes required storage. The storage may entail language-specific processing qualifiers.

Claims

1. A method for creating a diphone database useful for generating speech from text, the method including the steps of:

pronouncing a set of selected diphones;

processing each of the selected diphones for storage;

assigning each of the selected diphones to one or more of a plurality of different languages, at least a subset of the set of selected diphones being assigned to at least two of the different languages;

storing each of the selected diphones in a memory device; and

,

for each language, storing processing qualifiers for the diphones assigned to that language in the memory device in such a manner that when the diphones assigned to that language are accessed for the purpose of synthesizing speech in that language, the processing qualifiers specific to that language can be selectively extracted.

2. The method as set forth in claim 1, wherein the step of pronouncing is carried out by employing a single person to homophonically pronounce each of the selected diphones.

3. The method as set forth in claim 1, wherein the step of storing is carried out by storing each of the selected diphones in a separate, individually addressable storage location in the memory device.

4. The method as set forth in claim 1, wherein the processing qualifiers include diphone duration indicia.

5. The method as set forth in claim 1, wherein the processing qualifiers include intonation indicia.

6. The method as set forth in claim 1, wherein the processing qualifiers include aspiration indicia.

7. The method as set forth in claim 1, wherein the processing qualifiers include formant intensity and frequency indicia.

8. The method as set forth in claim 1, wherein the processing qualifiers include markers indicating the start and end of an aspiration sound.

9. The method as set forth in claim 1, wherein the processing qualifiers include pitch and rhythm indicia.

10. An apparatus for synthesizing speech in any selected one of a plurality of different languages from text, the apparatus including a diphone database created in accordance with the method set forth in claim 1.

11. The apparatus as set forth in claim 10, wherein the apparatus includes:

a processing section that receives the text in a form of diphone identifier strings and extracts the diphones identified by the diphone identifiers from the memory device, and that extracts the processing qualifiers from the memory device specific to the selected one of the plurality of different languages that is chosen by a user; and

,

transforming the text into speech in the selected one of the plurality of different languages using the diphones and processing qualifiers extracted by the processing section.

12. The apparatus as set forth in claim 10, wherein the apparatus includes:

processing means for receiving the text in a form of diphone identifier strings and extracting the diphones identified by the diphone identifiers from the memory device, and extracting the processing qualifiers from the memory device specific to the selected one of the plurality of different languages that is chosen by a user; and

,

transforming means for transforming the text into speech in the selected one of the plurality of different languages using the diphones and processing qualifiers extracted by the processing means.

13. A method for synthesizing speech in any selected one of a plurality of different languages from text by employing a diphone database created in accordance with the method of claim 1.

14. The method as set forth in claim 13, wherein the method includes the steps of:

receiving the text in a form of diphone identifier strings and extracting the diphones identified by the diphone identifiers from the memory device;

extracting the processing qualifiers from the memory device specific to the selected one of the plurality of different languages that is chosen by a user; and

,

transforming the text into speech in the selected one of the plurality of different languages using the diphones and processing qualifiers extracted by the processing means.