Text-to-speech system, text-to-speech method, and computer program product for synthesis modification based upon peculiar expressions

• US 9,570,067 B2
• Filed: 03/11/2015
• Issued: 02/14/2017
• Est. Priority Date: 03/19/2014
• Status: Active Grant

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation), Toshiba Digital Solutions Corporation (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Yamasaki, Tomohiro, Shimizu, Yuji, Yamanaka, Noriko, Yajima, Makoto, Miyamura, Yuichi
• Primary Examiner(s): Wozniak, James
• Application Number:
  US14/644,389
• Publication Number:
  US 20150269927A1
• Time in Patent Office: 706 Days
• Field of Search: 704/258, 704/260, 704/9
• US Class Current: 1/1
• CPC Class Codes: G10L 13/08 Text analysis or generation of parameters for speech synthesis out of text, e.g. grapheme to phoneme translation, prosody generation or stress or intonation determination
• CPC Class Codes: G10L 13/10 Prosody rules derived from text; Stress or intonation
• CPC Class Codes: G10L 2013/083 Special characters, e.g. punctuation marks

Abstract

According to an embodiment, a text-to-speech device includes a receiver to receive an input text containing a peculiar expression; a normalizer to normalize the input text based on a normalization rule in which the peculiar expression, a normal expression of the peculiar expression, and an expression style of the peculiar expression are associated, to generate normalized texts; a selector to perform language processing of each normalized text, and select a normalized text based on result of the language processing; a generator generate a series of phonetic parameters representing phonetic expression of the selected normalized text; a modifier modifies a phonetic parameter in the normalized text corresponding to the peculiar expression in the input text based on a phonetic parameter modification method according to the normalization rule of the peculiar expression; and a output unit to output a phonetic sound synthesized using the series of phonetic parameters including the modified phonetic parameter.

Claims

1. A text-to-speech system comprising a processing circuitry coupled to a memory, the processing circuit being configured to:

receive an input text which contains a peculiar expression representing an expression not used in normal expressions;

identify a position of the peculiar expression in the input text based on a normalization rule in which the peculiar expression, a normal expression for expressing the peculiar expression in a normal form, a non-linguistic expression style of the peculiar expression representing a manner in which the peculiar expression is read aloud, and a first cost are associated with one another, so as to generate one or more normalized texts;

calculate one or more combinations of one or more positions to which one or more normalization rules are to be applied;

calculate a total of the first cost or first costs in the case of applying the normalization rules for each combination of the combinations;

normalize the input text based on the normalization rules by using the combinations for which the total is smaller than a first threshold value;

perform language processing with respect to each of the normalized texts, and select a single normalized text based on result of the language processing;

generate a series of phonetic parameters representing phonetic expression of the single normalized text;

modify a phonetic parameter in the normalized text corresponding to the peculiar expression in the input text based on a phonetic parameter modification method according to the normalization rule of the peculiar expression; and

output a phonetic sound which is synthesized using the series of phonetic parameters including the modified phonetic parameter.

2. The system according to claim 1, wherein

the processing circuitry generates the series of phonetic parameters by selecting a synthesis unit from a synthesis unit dictionary, and

the processing circuitry modifies the synthesis unit, which is selected by the processing circuitry, based on a phonetic parameter modification method according to the normalization rule of the peculiar expression.

3. The system according to claim 1, wherein

the processing circuitry generates the series of phonetic parameters from an acoustic parameter based on a hidden Markov model, and

the processing circuitry modifies the acoustic parameter, which is selected by the processing circuitry, based on a phonetic parameter modification method according to the normalization rule of the peculiar expression.

4. The system according to claim 1, wherein the processing circuitry modifies the phonetic parameter so as to change the fundamental frequency of the phonetic sound output by the processing circuitry.

5. The system according to claim 1, wherein the processing circuitry modifies the phonetic parameter so as to change length of each sound included in the phonetic sound output by the processing circuitry.

6. The system according to claim 1, wherein the processing circuitry modifies the phonetic parameter so as to change pitch of the phonetic sound output by the processing circuitry.

7. The system according to claim 1, wherein the processing circuitry modifies the phonetic parameter so as to change volume of the phonetic sound output by the processing circuitry.

8. A text-to-speech method comprising:

receiving an input text which contains a peculiar expression representing an expression not used in normal expressions;

identifying a position of the peculiar expression in the input text based on a normalization rule in which the peculiar expression, a normal expression for expressing the peculiar expression in a normal form, and a non-linguistic expression style of the peculiar expression representing a manner in which the peculiar expression is read aloud, and a first cost are associated with one another, so as to generate one or more normalized texts;

calculating one or more combinations of one or more positions to which one or more normalization rules are to be applied;

calculating a total of the first cost or first costs in the case of applying the normalization rules for each combination of the combinations;

normalizing the input text based on the normalization rules by using the combinations for which the total is smaller than a first threshold value;

performing language processing with respect to each of the normalized texts, and selecting a single normalized text based on result of the language processing;

generating a series of phonetic parameters representing phonetic expression of the single normalized text;

modifying a phonetic parameter in the normalized text corresponding to the peculiar expression in the input text based on a phonetic parameter modification method according to the normalization rule of the peculiar expression; and

outputting a phonetic sound which is synthesized using the series of phonetic parameters including the modified phonetic parameter.

9. A computer program product comprising a non-transitory computer readable medium including programmed instructions, wherein the instructions, when executed by a computer, cause the computer to perform:

receiving an input text which contains a peculiar expression representing an expression not used in normal expressions;

identifying the position of the peculiar expression in the input text based on a normalization rule in which the peculiar expression, a normal expression for expressing the peculiar expression in a normal form, a non-linguistic expression style of the peculiar expression representing manner in which the peculiar expression is read aloud, and first cost are associated with one another, so as to generate one or more normalized texts;

calculating one or more combinations of one or more positions to which one or more normalization rules are to be applied;

calculating a total of the first cost or first costs in the case of applying the normalization rules for each combination of the combinations;

normalizing the input text based on the normalization rules by using the combinations for which the total is smaller than a first threshold value;

performing language processing with respect to each of the normalized texts, and selecting a single normalized text based on result of the language processing;

generating a series of phonetic parameters representing phonetic expression of the single normalized text;

modifying a phonetic parameter in the normalized text corresponding to the peculiar expression in the input text based on a phonetic parameter modification method according to the normalization rule of the peculiar expression; and

outputting a phonetic sound which is synthesized using the series of phonetic parameters including the modified phonetic parameter.