Method and system for preselection of suitable units for concatenative speech

• US 7,460,997 B1
• Filed: 08/22/2006
• Issued: 12/02/2008
• Est. Priority Date: 06/30/2000
• Status: Expired due to Fees

• Current Assignee: Cerence Operating Company (Cerence Inc.)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Conkie, Alistair D.
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US11/466,229
• Time in Patent Office: 833 Days
• Field of Search: 704/258, 704/259, 704/260, 704/263, 704/266
• US Class Current: 704/258
• CPC Class Codes: G10L 13/07 Concatenation rules
• CPC Class Codes: G10L 2015/022 Demisyllables, biphones or triphones being the recognition units

Abstract

A system and method for improving the response time of text-to-speech synthesis utilizes “triphone contexts” (i.e., triplets comprising a central phoneme and its immediate context) as the basic unit, instead of performing phoneme-by-phoneme synthesis. The method comprises a method of generating a triphone preselection cost database for use in speech synthesis, the method comprising 1) selecting a triphone sequence u₁-u₂-u₃, 2) calculating a preselection cost for each 5-phoneme sequence u_a-u₁-u₂-u₃-u_b, where u₂ is allowed to match any identically labeled phoneme in a database and the units u_a and u_b vary over the entire phoneme universe and 3) storing a group of the selected triphone sequences exhibiting the lowest costs in a triphone preselection cost database.

Claims

1. A computing device that generates a database for use in speech synthesis, the computing device generating the database according to a method comprising:

selecting a triphone sequence;

calculating a preselection cost for each 5-phoneme sequence where a unit of the 5-phoneme sequence is allowed to match any identically labeled phoneme in a database and at least two units of the 5-phoneme sequence vary over the entire phoneme universe; and

storing a group of the selected triphone sequences exhibiting the lowest costs in a triphone preselection cost database by;

determining a plurality of N least cost database units for the particular 5-phoneme context;

performing the union of the N least cost units for all combinations of the at least two units;

storing the union created in the step of performing the union in the triphone preselection cost database; and

repeating steps of selecting, calculating and storing a group of the selected triphone sequences for each possible triphone sequence.

2. The computing device of claim 1, wherein the triphone sequence comprises u₁-u₂-u₃ and the 5-phoneme sequence comprises u_a-u₁-u₂-u₃-u_b, where u₂ is allowed to match any identically labeled phoneme in a database and the at least two units comprise u_a and u_b vary over the entire phoneme universe.

3. The computing device of claim 1, wherein the method for generating the database further comprises generating a key to index each triphone in the database.

4. The computing device of claim 1, wherein a plurality of fifty least costs sequences for any possible 5-phone context are stored.

5. The computing device of claim 1 wherein the preselection cost is the target cost or an element of the target cost.

6. A method for generating a triphone preselection cost database for use in speech synthesis, the method comprising, for each of plurality of triphone sequences:

calculating a preselection cost for each 5-phoneme sequence, wherein a triphone sequence of the plurality of triphone sequences is included in each 5-phoneme sequence;

storing a group of triphone sequences exhibiting the lowest costs in a triphone preselection cost database by;

a) determining a plurality of N least cost database units for the particular 5-phoneme context;

b) performing the union of the N least cost units for all combinations of two selected units from the 5-phoneme sequence; and

c) storing the union created in step b) in the triphone preselection cost database.

7. The method of claim 6, wherein the selected triphone sequence comprises u₁-u₂-u₃ and the 5-phoneme sequence comprises u_a-u₁-u₂-u₃-u_b, where u₂ is allowed to match any identically labeled phoneme in a database and the units u_a and u_b vary over the entire phoneme universe, and wherein the two selected units from the 5-phoneme sequence for which the union is performed are u_a and u_b.

8. The method of claim 6, further comprising generating a key to index each triphone in the triphone cost selection database.

9. The method of claim 6, wherein a plurality of fifty least costs sequences for any possible 5-phoneme context are stored.

10. The method of claim 6, wherein the preselection cost is a target cost or an element of the target cost.

11. A computer-readable medium storing instructions for controlling a computing device to generate a triphone preselection cost database for use in speech synthesis, the instructions comprising, for each of plurality of triphone sequences:

calculating a preselection cost for each 5-phoneme sequence, wherein a triphone sequence of the plurality of triphone sequences is included in each 5-phoneme sequence;

storing a group of triphone sequences exhibiting the lowest costs in a triphone preselection cost database by;

a) determining a plurality of N least cost database units for the particular 5-phoneme context;

b) performing the union of the N least cost units for all combinations of two selected units from the 5-phoneme sequence; and

c) storing the union created in step b) in the triphone preselection cost database.

12. The computer-readable medium of claim 11, wherein the selected triphone sequence comprises u₁-u₂-u₃ and the 5-phoneme sequence comprises u_a-u₁-u₂-u₃-u_b, where u₂ is allowed to match any identically labeled phoneme in a database and the units u_a and u_b vary over the entire phoneme universe, and wherein the two selected units from the 5-phoneme sequence for which the union is performed are u_a and u_b.

13. The computer-readable medium of claim 11, further comprising generating a key to index each triphone in the triphone cost selection database.

14. The computer-readable medium of claim 11, wherein a plurality of fifty least costs sequences for any possible 5-phoneme context are stored.

15. The computer-readable medium of claim 11, wherein the preselection cost is a target cost or an element of the target cost.