Voice synthesizing apparatus using database having different pitches for each phoneme represented by same phoneme symbol

• US 7,065,489 B2
• Filed: 03/08/2002
• Issued: 06/20/2006
• Est. Priority Date: 03/09/2001
• Status: Active Grant

• Current Assignee: Yamaha Corporation
• Original Assignee: Yamaha Corporation
• Inventors: Hisaminato, Yuji, Sanjaume, Jordi Bonada
• Primary Examiner(s): Azad, Abul K.
• Application Number:
  US10/094,154
• Publication Number:
  US 20020184032A1
• Time in Patent Office: 1,565 Days
• Field of Search: 704258-270
• US Class Current: 704/268
• CPC Class Codes: G10L 13/033 Voice editing, e.g. manipulating the voice of the synthesiser
• CPC Class Codes: G10L 13/06 Elementary speech units used in speech synthesisers; Concatenation rules

Abstract

A voice synthesizing apparatus comprises: a memory that stores phoneme pieces having a plurality of different pitches for each phoneme represented by a same phoneme symbol; a reading device that reads a phoneme piece by using a pitch as an index; and a voice synthesizer that synthesizes a voice in accordance with the read phoneme piece.

Claims

1. A voice synthesizing apparatus comprising:

a timbre storing device that stores voice feature parameters of a plurality of phoneme, each parameter having a plurality of different pitches for each phoneme represented by a same phoneme symbol and being indexed by a phoneme name and a pitch;

a phoneme template storing device that stores a plurality of templates each having a sequence of feature parameters disposed at a predetermined time interval and being indexed by a phoneme name and a pitch, the templates including a stationary template derived from voices having stable phonemes and an articulation template derived from voices in a concatenated part of the phonemes;

a note template storing device that stores a plurality of templates each having a sequence of feature parameters disposed at a predetermined time interval and being indexed by a phoneme name and a pitch, the templates including at least a note attack template having feature parameters in a voice rising part and a note-to-note template having feature parameters in a pitch changing part;

a reading device that reads the feature parameter from the timbre storing device and the templates from the phoneme template storing device and the note template storing device by using information regarding the phoneme and a pitch of a voice to be synthesized changing over time as indices; and

a voice synthesizer that synthesizes a voice in accordance with the read feature parameter added with the templates read from the phoneme template storing device and the note template storing device.

2. A voice synthesizing apparatus according to claim 1, wherein the templates stored in the note templates storing device include a note release template having feature parameters in a voice falling part.

3. A voice synthesizing apparatus according to claim 1, wherein each feature parameter in the templates is stored by a differential value.

4. A voice synthesizing apparatus according to claim 1, further including a calculator that calculates a voice feature parameter matching a pitch of the voice to be synthesized by interpolation, when the voice feature parameter matching a pitch of the voice to be synthesized is not stored in the timbre storing device.

5. A voice synthesizing apparatus according to claim 1, wherein the articulation template is lineally stretched.

6. A voice synthesizing apparatus according to claim 1, wherein the reading device reads the note-to-note template in accordance with an added value of a weighted change amount of frequencies and an average value of start pitches and end pitches.

7. A voice synthesizing apparatus according to claim 1, wherein the feature parameters further is indexed by dynamics.

8. A voice synthesizing apparatus according to claim 1, wherein the feature parameters further is indexed by a lip opening value.

9. A voice synthesizing method comprising:

reading a feature parameter, by using as indices information regarding a phoneme and a pitch of a voice to be synthesized, from a timbre storing means which stores voice feature parameters of a plurality of phoneme, each parameter having a plurality of different pitches for each phoneme represented by a same phoneme symbol, and the feature parameter being indexed by a phoneme name and a pitch;

reading a template, by using as indices information regarding a phoneme and a pitch of a voice to be synthesized changing over time, from a phoneme template storing means which stores a plurality of templates, each template having a sequence of feature parameters disposed at a predetermined time interval and being indexed by a phoneme name and a pitch, the templates including a stationary template derived from voices having stable phonemes and an articulation template derived from voices in a concatenated part of the phonemes;

reading a template, by using as indices information regarding a phoneme and a pitch of a voice to be synthesized, from a note template storing means which stores a plurality of templates, each template having a sequence of feature parameters disposed at a predetermined time interval and being indexed by a phoneme name and a pitch, the templates including at least a note attack template having feature parameters in a voice rising part and a note-to-note template having feature parameters in a pitch changing part; and

synthesizing a voice in accordance with the read feature parameter added with the templates read from the phoneme template storing means and the note template storing means.

10. A computer-readable storage medium having encoded thereon, program code including instructions which when executed cause:

reading a feature parameter, by using as indices information regarding a phoneme and a pitch of a voice to be synthesized, from a timbre storing means which stores voice feature parameters of a plurality of phoneme, each parameter having a plurality of different pitches for each phoneme represented by a same phoneme symbol, and the feature parameter being indexed by a phoneme name and a pitch;

reading a template, by using as indices information regarding a phoneme and a pitch of a voice to be synthesized changing over time, from a phoneme template storing means which stores a plurality of templates, each template having a sequence of feature parameters disposed at a predetermined time interval and being indexed by a phoneme name and a pitch, the templates including a stationary template derived from voices having stable phonemes and an articulation template derived from voices in a concatenated part of the phonemes;

reading a template, by using as indices information regarding a phoneme and a pitch of a voice to be synthesized, from a note template storing means which stores a plurality of templates, each template having a sequence of feature parameters disposed at a predetermined time interval and being indexed by a phoneme name and a pitch, the templates including at least a note attack template having feature parameters in a voice rising part and a note-to-note template having feature parameters in a pitch changing part; and

synthesizing a voice in accordance with the read feature parameter added with the templates read from the phoneme template storing means and the note template storing means.