Speech information processing method and apparatus and storage medium using a segment pitch pattern model

• US 6,826,531 B2
• Filed: 03/28/2001
• Issued: 11/30/2004
• Est. Priority Date: 03/31/2000
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Fukada, Toshiaki
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US09/818,599
• Publication Number:
  US 20010032078A1
• Time in Patent Office: 1,343 Days
• Field of Search: 704/258, 704/260, 704/268, 704/269
• US Class Current: 704/258
• CPC Class Codes: G10L 13/10 Prosody rules derived from text; Stress or intonation
• CPC Class Codes: G10L 15/1807 using prosody or stress

Abstract

A speech information processing apparatus synthesizes speech with natural intonation by modeling time change in fundamental frequency of a predetermined unit of phoneme. When a predetermined unit of phonological series is inputted, fundamental frequencies of respective phonemes constructing the phonological series are generated based on a segment pitch pattern model. Phonemes are synthesized based on the generated fundamental frequencies of the respective phonemes.

Claims

1. A speech information processing method comprising:

an input step of inputting a predetermined unit of phonological series;

a generation step of generating fundamental frequencies of respective phonemes constructing said phonological series based on a segment pitch pattern model; and

a speech synthesis step of synthesizing speech based on said fundamental frequencies of the respective phonemes generated at said generation step.

2. The speech information processing method according to claim 1, wherein said segment pitch pattern model is a model based on at least any one of a phoneme, a syllable and a word as a unit.

3. The speech information processing method according to claim 1, wherein said segment pitch pattern model is obtained by modeling time change in a fundamental frequency of a phoneme belonging to a predetermined phonemic environment with a polynomial segment model.

4. The speech information processing method according to claim 1, wherein said segment pitch pattern model is a model in consideration of at least one of an accent type, the number of moras, a mora position and a part of speech.

5. The speech information processing method according to claim 1, wherein said segment pitch pattern model is obtained by modeling with at least any one of single mixed distribution and multiple mixed distribution.

6. The speech information processing method according to claim 1, wherein said segment pitch pattern model is normalized by at least any one of an accent phrase, a word, a phrase and a sentence as a unit.

7. A computer-readable storage medium holding a program for executing the speech information processing method in claim 1.

8. A speech information processing apparatus comprising:

input means for inputting a predetermined unit of phonological series;

generation means for generating fundamental frequencies of respective phonemes constructing said phonological series based on a segment pitch pattern model; and

speech synthesis means for synthesizing speech based on said fundamental frequencies of the respective phonemes generated by said generation means.

9. The speech information processing apparatus according to claim 8, wherein said segment pitch pattern model is a model based on at least any one of a phoneme, a syllable and a word as a unit.

10. The speech information processing apparatus according to claim 8, wherein said segment pitch pattern model is obtained by modeling time change in a fundamental frequency of a phoneme belonging to a predetermined phonemic environment with a polynomial segment model.

11. The speech information processing apparatus according to claim 10, wherein said segment pitch pattern model is a model in consideration of at least one of an accent type, the number of moras, a mora position and a part of speech.

12. The speech information processing apparatus according to claim 10, wherein said segment pitch pattern model is obtained by modeling with at least any one of single mixed distribution and multiple mixed distribution.

13. The speech information processing method according to claim 8, wherein said segment pitch pattern model is normalized by at least any one of an accent phrase, a word, a phrase and a sentence as a unit.