JP3513071B2 - Speech synthesis method and speech synthesis device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice synthesizing method and a voice synthesizing apparatus for synthesizing text voices.

[0002]

2. Description of the Related Art Artificially producing a voice signal from an arbitrary sentence is called text-to-speech synthesis. Usually, a text synthesis system is composed of three stages: text analysis, synthesis parameter generation, and speech synthesis.

FIG. 10 shows an example of the configuration of a conventional general speech synthesizer. As shown in FIG. 10, the conventional general text synthesizing system normally uses the text analysis unit 10.
10, synthesis parameter generation unit 1001, speech synthesis unit 10
13 and a prosody control dictionary 1002. The input text 1102 is first analyzed by the text analysis unit 1010.
In, morphological analysis and syntactic analysis are performed, and language information 1104 is output. The language information 1104 includes various information necessary for generating synthetic parameters, such as a phonetic symbol string corresponding to the reading of the text 1102, information on an accent phrase as a unit of prosody control, an accent position, and a part of speech. Next, the synthesis parameter generation unit 1001 performs prosody control based on the language information 104 with reference to the prosody control dictionary 1002, and generates the synthesis parameter 1100.
The synthesis parameter 1100 is composed of prosodic parameters such as fundamental frequency, phoneme duration and power, and phoneme parameters such as phoneme symbol strings. Then, the voice synthesis unit 10
13 generates voice information 1108 according to the phoneme information and prosody information designated by the synthesis parameter 1100.

In such a synthesizing system, it is usual to synthesize a voice having a tone (so-called reading tone) as when a human reads a sentence, but in recent years, various synthesizing voices have been controlled by controlling the utterance style. A method of generating is proposed. For example, Japanese Unexamined Patent Publication No. 10-11083 discloses a method of controlling the utterance style of synthetic speech using a prosody control dictionary of a plurality of utterance styles including a reference utterance style (reading tone, etc.). FIG. 11 shows the configuration of this conventional speech synthesizer.

The difference from the above-mentioned general text synthesizing system is that it has a plurality of prosodic control dictionaries (201 in FIG. 11).
4, 2015), utterance style designation information 2103
Prosodic control dictionary selected based on (eg 2015)
And the reference utterance style prosody control dictionary 2016 are used to synthesize parameters 2105 and 2 respectively.
106 is generated, and the utterance style emphasizing unit 2012 corrects the utterance style in accordance with the emphasis degree designation information 2101. Plural prosody control information 2014, 2015
Is a prosody control dictionary of an utterance style different from the reference utterance style, and includes, for example, a conversation style and an announcer style. The utterance style emphasizing unit 2012 calculates the difference between the prosody parameters of the utterance style synthesis parameter 2105 selected by the utterance style designation information 2103 and the reference utterance style synthesis parameter 2106, and obtains the emphasis degree designation information 2101 and the difference. By correcting the prosody parameter of the synthesis parameter 2106 accordingly, the synthesis parameter 210 in which the utterance style is adjusted
7 is generated.

[0006]

As described above, in the conventional speech synthesis method, it is only possible to change to an intermediate utterance style between the reference utterance style and the selected one utterance style, and it is always The utterance style is constant and there is little freedom to change. In addition, you can change
Utterance style (reading, conversation, announcer, etc.)
However, there is a problem that it is impossible to change the speaker's individuality (voice of Mr. A, voice of Mr. B, etc.) and emotion (angry voice, sad voice, etc.).

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a voice synthesizing method and a voice synthesizing apparatus capable of improving the diversity of prosody of synthesized voice by text voice synthesis. .

[0008]

SUMMARY OF THE INVENTION According to the present invention, a prosodic parameter generated by using a plurality of prosodic control dictionaries is interpolated at an arbitrary ratio to generate a prosodic parameter. It can be generated. Here, the plurality of prosodic control dictionaries may have different utterance styles, different personality, different age / sex, or different emotions. You may use what has various characteristics by the combination of.

That is, the speech synthesis method according to the present invention (Claim 1) uses the plurality of prosody control dictionaries to generate the first prosody parameters according to the input linguistic information, and the prosody control dictionaries are generated for each of the prosody control dictionaries. According to the designated weight information, interpolation processing is performed between the plurality of first prosody parameters to generate second prosody parameters, and synthetic speech is generated according to the second prosody parameters.

Here, the linguistic information is obtained by analyzing text such as a syllable symbol string corresponding to reading of text, information of accent phrase which is a unit of prosody control, position of accent, part of speech, catching, etc. The information you get,
It is composed of additional information that specifies the average rate of occurrence and the volume of voice.

Here, the prosodic control dictionary is referred to in order to control prosody such as fundamental frequency, phoneme duration, power, and pause of synthesized speech. For example, a typical variation pattern of fundamental frequency. Also, statistical model parameters of control quantities such as accent components, phoneme durations, powers, and pause lengths, or rules represented by decision trees can be considered.

Here, the prosody parameter is a set of parameters that characterize the prosody of the synthetic speech such as the fundamental frequency, the phoneme duration, the power, and the pause.

Here, the interpolation process between the prosody parameters is a process of generating an intermediate prosody parameter of the plurality of prosody parameters by a process such as a weighted average among the plurality of prosody parameters. However, the interpolation processing here includes so-called extrapolation processing in which the weight becomes negative, and in this case, the generated prosody parameter is an intermediate one of the plurality of prosody parameters. However, the characteristics of one of the prosody parameters may be more emphasized. Further, this interpolation processing may be performed on all the prosody parameters, or may be performed on some parameters, for example, only the fundamental frequency. In addition, for example, the fundamental frequency and the phoneme duration may have different weights for interpolation.

Further, preferably, the weight information may be changed in the sentence.

In the speech synthesis method according to the present invention (claim 3), interpolation processing is performed between a plurality of first prosody control dictionaries to generate a second prosody control dictionary, and the inputted language information According to the above, a prosody parameter is generated using the second prosody control dictionary, and a synthetic speech is generated according to the prosody parameter.

Here, the interpolating process between the prosodic dictionaries is a phonological unit having an intermediate characteristic of the plural prosodic control dictionaries by a process such as a weighted average among corresponding information in the plural prosodic control dictionaries. This is a process of generating a control dictionary. However, similar to the above-described interpolation of the prosody parameters, the interpolation processing here includes so-called extrapolation processing in which the weight becomes negative. In this case, the generated prosody control dictionary The characteristic is not intermediate between the prosodic control dictionaries, and may be a characteristic in which one of the prosodic control dictionaries is more emphasized. Further, this interpolation processing may be performed on the entire prosody control dictionary, or may be performed only on a part thereof, for example, only a portion related to fundamental frequency control. Further, for example, the weight for interpolation of the part related to the fundamental frequency may be different from the weight for interpolation of the part related to phoneme duration control.

Further, the speech synthesis method according to the present invention (claim 4) uses a second prosody control dictionary generated by performing an interpolation process among a plurality of first prosody control dictionaries,
Generate prosody parameters according to the input linguistic information,
It is characterized in that synthetic speech is generated according to the prosody parameter.

Further, preferably, the prosody control dictionary is a representative pattern representing a typical fundamental frequency change pattern,
Alternatively, information having equivalent information, such as a typical pitch period change pattern, may be included.

Further, the speech synthesizer according to the present invention (claim 6) includes means for generating a first prosody parameter using a plurality of prosody control dictionaries in accordance with the input linguistic information, and the prosody control dictionary. Means for interpolating between a plurality of the first prosody parameters according to the weight information designated for each to generate a second prosody parameter; and means for generating a synthetic speech according to the second prosody parameter. It is characterized by having.

Further, the speech synthesizer according to the present invention (claim 7) is input with a means for performing interpolation processing between a plurality of first prosody control dictionaries to generate a second prosody control dictionary. It is characterized by further comprising means for generating a prosody parameter using the second prosody control dictionary according to language information and means for generating a synthetic voice according to the prosody parameter.

Further, in the prosody control dictionary creating method according to the present invention, the weight information designated for each of the plurality of prosody control dictionaries is input, and in accordance with the inputted weight information, among the plurality of first prosody control dictionaries. The second prosody control dictionary is generated by performing interpolation processing.

Further, the prosody control dictionary creating apparatus according to the present invention has means for inputting weight information designated for each of the plurality of prosody control dictionaries, and a plurality of first prosody control dictionaries according to the inputted weight information. Means for performing interpolating processing between them to generate a second prosody control dictionary.

According to the present invention (claim 10), the first prosody parameter is generated using a plurality of prosody control dictionaries in accordance with the input linguistic information, and the weight information designated for each prosody control dictionaries is generated. A computer-readable record of a program for performing an interpolation process between a plurality of the first prosody parameters to generate a second prosody parameter, and generating a synthetic speech according to the second prosody parameter. It is a medium.

According to the present invention (claim 11), a second prosody control dictionary is generated by performing an interpolating process among a plurality of first prosody control dictionaries, and the second prosody control dictionary is generated according to the input language information. Generate prosody parameters using the prosody control dictionary of
It is a computer-readable recording medium in which a program for generating synthetic speech according to the prosody parameter is recorded.

It should be noted that the present invention related to the apparatus is established as an invention related to the method, and the present invention related to the method is also established as an invention related to the apparatus.

Further, the present invention related to an apparatus or method is for causing a computer to execute a procedure corresponding to the present invention (or for causing a computer to function as means corresponding to the present invention, or for a computer corresponding to the present invention. It also holds as a computer-readable recording medium in which a program (for realizing the function to perform) is recorded.

According to the present invention, prosodic parameters generated by using a plurality of prosodic control dictionaries are interpolated with arbitrary weights to generate prosodic parameters, and speech synthesis is performed using the prosodic parameters. , It is possible to generate synthetic speech having various prosodic features.

Further, according to the present invention, the calculation amount is increased by performing interpolation processing on a plurality of prosody control dictionaries in advance to generate a prosody control dictionary and performing speech synthesis using the prosody control dictionaries. It is possible to generate synthetic speech having a wide variety of prosodic features.

[0029]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

(First Embodiment) FIG. 1 shows a first embodiment of the present invention.
3 is a block diagram showing a configuration example of a voice synthesizing device (or voice synthesizing software) for realizing the voice synthesizing method according to the embodiment of FIG.

As shown in FIG. 1, this speech synthesizer comprises a text analysis section 10, a synthesis parameter generation section 20,
A synthesis parameter interpolation unit 22, a voice synthesis unit 13, and a plurality of prosody control dictionaries (three of 24 to 26 in FIG. 1) are provided.

The basic configuration and operation of each unit are as follows (basically, the text analysis section 10, the synthesis parameter generation section 20, the synthesis parameter interpolation section 22, and the speech synthesis section 13 are processed in this order. Is done).

The text analysis unit 10 performs morphological analysis and syntactic analysis on the input text 102,
The language information 104 is generated. The language information 104 includes various information necessary for generating a synthesis parameter, such as a phonetic symbol string corresponding to reading of the text 102, information on an accent phrase serving as a unit of prosody control, an accent position, and a part of speech.

The synthesis parameter generation unit 20 uses the language information 1
04, the synthesis parameter 204 is generated by referring to the prosody control dictionary 24, the synthesis parameter 205 is generated by referring to the prosody control dictionary 25, and the synthesis parameter 206 is generated by referring to the prosody control dictionary 26.

The synthesis parameter interpolation unit 22 uses the weight information 2
In accordance with 01, interpolation processing is performed on the prosody parameters of the synthesis parameter 204, the synthesis parameter 205, and the synthesis parameter 206 to generate the synthesis parameter 207.

The voice synthesizing unit 13 synthesizes parameters 207.
Voice information 1 according to the phonological information and prosody information specified in
08 is generated.

In the following, the operation example of this embodiment will be described in detail by using the fundamental frequency as an example of the prosody parameter.

The language information 104 by the text analysis unit 10
Is generated, the synthesis parameter generation unit 20 generates a plurality of synthesis parameters by referring to a plurality of prosody control dictionaries according to the language information 104.

FIG. 2 is a functional block diagram showing the processing for generating the fundamental frequency in the synthesis parameter generating section 20. Note that FIG. 2 shows one prosody control dictionary (actually, the process of FIG. 2 is performed for each of the plurality of prosody control dictionaries).

The fundamental frequency control dictionary 48 is a part of the prosody control dictionary, and is composed of a representative pattern dictionary 45, a representative pattern selection rule 46, and an offset generation rule 47.
The representative pattern dictionary 45 is a set of typical basic frequency change patterns in units of accent phrases, and stores patterns such as that shown in FIG. 3, for example. The representative pattern selection unit 41 refers to the representative pattern selection rule 46 according to the language information 104, and selects the representative pattern 401 expected to be most suitable from the representative pattern dictionary for each accent phrase. The offset generation unit 44 refers to the offset generation rule 47 according to the language information 104 and generates an offset 404 that specifies the average height of the accent phrase for each accent phrase. The offset processing unit 42 moves the representative pattern dictionary 45 in parallel by the offset 404 on the logarithmic frequency axis to generate the basic frequency pattern 4 for each accent.
02 is generated. The pattern connection unit 43 smoothly connects the basic frequency patterns 402 for each accent phrase and outputs the basic frequency pattern 403 of the entire sentence. Taking the text "I am currently testing the microphone" as an example, a representative pattern 401, a fundamental frequency pattern 402, and a fundamental frequency pattern 403 are shown in FIGS. 4 (a), (b), and (c), respectively.

The synthesis parameter generation unit 20 performs the above-described processing of fundamental frequency pattern generation in a plurality of prosody control dictionaries 2.
4.25.26 to generate a fundamental frequency pattern, and synthesize parameters 204.205.20.
6 is output.

Subsequently, the synthesis parameter interpolation unit 22 performs a weighted average process on the fundamental frequency patterns of the synthesis parameters 204, 205 and 206 according to the weight information 201 to generate a fundamental frequency pattern, and the synthesis parameter 2
07 is output.

The weight information 201 is represented by a set of n (three in the case of FIG. 1) weighting factors corresponding to n (n is a plurality) prosodic control dictionaries (24 to 26 in FIG. 1). To be done. Figure 5
Shows an example of an input means (for example, by GUI) of the weight information 201. In this example, the three prosodic control dictionaries 24, 25, and 26 correspond to prosodic control dictionaries expressing the prosodic features of the tone of Mr. A, Mr. B, and Mr. C, respectively. The position of the pointer indicated by a black circle specifies which prosody to which the prosody is similar. The position of the pointer is
The values of the weight information 201 for a, b, c and d in FIG. 5 are shown in FIG. If the position of the pointer is a, the weight information 20
1 has a prosody intermediate to those of 3 people, and in the case of b, B
The prosody most resembles that of Mr. C, and the features of Mr. A and Mr. A are slightly included. In the case of c, the prosody is intermediate between Mr. A and Mr. C, and the feature of Mr. B is not included, and in the case of d, Is a prosody that exaggerates the characteristics of Mr. A's prosody. In this way, by controlling the weight information, it becomes possible to generate synthetic speech with various personalities.

As another configuration example of the synthesis parameter interpolation unit 22, it is possible to change the weight information in the sentence. FIG. 7 shows an example of weight information that changes according to the text in response to "I'm in this lottery, I can play and live for the rest of my life." In this example, the three prosody control dictionaries 24, 25, and 26 correspond to prosody of the same person when the person is calm, surprised, or pleased. By performing the interpolation process according to such changing weight information, it becomes possible to generate a fundamental frequency pattern expressing a fine change in emotion.

The weight information 201 is input by the user, given by another program (process),
Various forms are possible, such as a form provided for each predetermined unit of text (for example, a sentence unit, a sentence component unit), a form generated by analyzing the text by the text analysis unit 10.

In the present embodiment, the basic frequency control model based on the representative pattern is used, but various basic frequency control models such as a model that approximates a pattern with a function, such as the so-called Fujisaki model, may be used. It can be used.

In the present embodiment, the fundamental frequency has been described as an example of the prosody parameter, but the prosody parameters such as prosody duration, power, and pause can be implemented in the same manner. That is, sequences of phoneme duration, power, pause, etc. are generated using a plurality of prosodic control dictionaries, and interpolation processing is performed according to the weight information described above to generate synthetic speech having various prosodic features. Can be generated.

As described above, according to the present embodiment, the prosody parameters generated using a plurality of prosody control dictionaries are interpolated with arbitrary weights to generate prosody parameters, and the prosody parameters are By performing speech synthesis by using it, it is possible to generate synthetic speech having various prosodic features.

(Second Embodiment) Next, a second embodiment of the present invention will be described.

In the first embodiment, interpolation processing is performed on the prosody parameters generated by referring to a plurality of prosody control dictionaries, whereas in the second embodiment, a plurality of prosody control dictionaries are used. The difference is that interpolation processing is performed in advance to create a prosody control dictionary in which the prosody features are adjusted.

FIG. 8 is a block diagram showing a configuration example of a voice synthesizing apparatus (or voice synthesizing software) for realizing the voice synthesizing method according to this embodiment.

As shown in FIG. 8, this speech synthesizer comprises a text analysis unit 10, a synthesis parameter generation unit 30,
The prosody control dictionary interpolating unit 31, the voice synthesizing unit 13, and a plurality of prosody control dictionaries (3 of 24 to 26 in FIG.
And a prosody control dictionary 32 obtained by interpolation.

In the following, the parts different from the first embodiment will be mainly described.

In this embodiment, the prosody control dictionary interpolation processing and the speech synthesis processing are roughly divided into two. That is, the prosody control dictionary interpolating unit 31 generates the prosody control dictionary 32 based on the plurality of prosody control dictionaries (24 to 26) in advance.
After that, using the prosody control dictionary 32, the text analysis unit 1
0. The synthesis parameter generation unit 30 and the voice synthesis unit 13 perform voice synthesis. If the content of the prosody control dictionary 32 is desired to be modified or changed, the prosody control dictionary interpolating unit 31 generates the prosody control dictionary 32 again.

Since the configuration and operation of the text analysis unit 10, the synthesis parameter generation unit 30, and the voice synthesis unit 13 are basically the same as those in FIG. 10, the description of each unit will be omitted here.

The prosody control dictionary interpolating unit 31 will be described below.

The prosody control dictionary interpolation unit 31 performs interpolation processing on the plurality of prosody control dictionaries 24, 25, and 26 according to the weight information 301 to generate a prosody control dictionary 32 (the synthesis parameter generation unit 30 , According to the language information 104,
The synthesis parameter 305 is referred to by referring to the prosody control dictionary 32.
Will be generated).

In the following, the operation example of this embodiment will be described in detail by using the fundamental frequency as an example of the prosody parameter.

The fundamental frequency control model in the synthesis parameter generation unit 30 will be described using the model based on the representative pattern described in FIG. 2 as in the first embodiment.

In this case, the basic frequency control dictionary is composed of a representative pattern dictionary, a representative pattern selection rule, and an offset generation rule. However, the representative pattern selection rule is
It is assumed that the prosody control dictionary is created so that all the prosody control dictionaries 24, 25, and 26 are common.
In this case, the representative pattern selection rule of the prosody control dictionary 32 may be a copy of one of the representative pattern selection rules of the plurality of prosody control dictionaries 24, 25, and 26. Therefore, the processing in the prosody control dictionary interpolating unit 31 interpolates the representative pattern dictionaries of the plurality of prosody control dictionaries 24, 25, and 26 to generate the representative pattern dictionary of the prosody control dictionary 32.
The offset generation rules of the prosody control dictionaries 24, 25, and 26 are interpolated to generate the offset generation rules of the prosody control dictionary 32.

First, the interpolation process of the representative pattern dictionary will be described.

Each representative pattern dictionary is composed of N representative patterns. For the representative patterns No. 1 to N, the representative patterns having the same number in each representative pattern dictionary are weighted and averaged according to the weight information 301 to generate the interpolated representative patterns.

Next, the interpolation process of the offset generation rule will be described.

The offset generation rule can be performed by using a quantification type I which is one of statistical models. FIG. 9 shows an example of the offset generation rule based on the quantification type I. The offset value is given by the sum of the coefficient corresponding to the category to which each language information belongs and the average value m. For example, when the position of a certain accent phrase is in the sentence, the number of moras is 4, and the part of speech is a noun, the offset value is m + a ₂ + b ₄ + c ₁ . Therefore, the offset generation rule in which the interpolation processing is performed is performed by performing the weighted average processing on the coefficients corresponding to the same category of each offset generation rule according to the weight information 301.

The weight information 301 is represented by a set of n (three in the case of FIG. 8) weighting factors corresponding to n (n is a plurality) prosodic control dictionaries (24 to 26 in the case of FIG. 8). And generate a prosody control dictionary by changing the weighting factors,
Similar to the first embodiment, it is possible to generate synthetic speech having various prosodic features.

Thereafter, using the prosody control dictionary 32, the text analysis unit 10, the synthesis parameter generation unit 30, and the voice synthesis unit 13 can perform voice synthesis.

It is also possible to prepare a plurality of types of prosody control dictionaries 32 generated by making the weight information 301 different, and select and use them appropriately.

Although the basic frequency control model based on the representative pattern is used in the present embodiment, various basic frequency control models such as the so-called Fujisaki model which approximates the pattern by function are used. It is possible.

Further, in the present embodiment, the fundamental frequency control is explained as an example of the prosodic parameter, but the prosodic parameter such as the phoneme duration, the power, and the pause can also be implemented in the same form. That is, the phoneme duration, power, pause, etc. are quantified by I
Since it is possible to control using a statistical model such as a class, the prosodic control with various prosodic characteristics is performed by interpolating the model parameters according to the weight information as in the offset generation rule described above. Information can be generated.

In the present embodiment, since the prosody parameters are generated only by using the prosody control dictionary generated by performing the interpolation process on a plurality of prosody control dictionaries in advance, the prosody parameter generation at the time of speech synthesis is performed. There is an advantage that the amount of calculation is smaller than that of the conventional speech synthesis method.

By the way, in the present embodiment, the text analysis unit 10 and the synthesis parameter generation unit 3 among the components shown in FIG.
It is also possible to adopt a configuration in which it is configured as a voice synthesizing device (or voice synthesizing software) including 0, the voice synthesizing unit 13, and the prosody control dictionary 32. Alternatively, it may be configured as a voice synthesis device (or voice synthesis software) including the text analysis unit 10, the synthesis parameter generation unit 30, and the voice synthesis unit 13, and the prosody control dictionary 32 may be input separately. With such a configuration, for example, a prosody control dictionary 32 suitable for the tone of each character of a video game is created in advance, and the character is used in a game in which a synthetic voice is spoken in the game. It can be applied to devices or application programs.

Similarly, the prosody control dictionary interpolating unit 31 and the prosody control dictionary 32 may be configured as a prosody control dictionary creating device (or prosody control dictionary creating software) including a plurality of prosody control dictionaries as materials. . Alternatively, the prosody control dictionary creation device (or the prosody control dictionary creation software) including the prosody control dictionary interpolating unit 31 may be configured, and a plurality of material prosody control dictionaries as materials may be input separately. With such a configuration, various prosody control dictionaries 32
Can be created and used by the user himself or can be created and provided to the user by the manufacturer.

As described above, according to the present embodiment, by interpolating a plurality of prosody control dictionaries in advance to generate a prosody control dictionary, and performing speech synthesis using the prosody control dictionaries. , It is possible to generate synthetic speech having various prosodic features without increasing the amount of calculation.

Each of the above functions can be realized as hardware or software.

Further, the present embodiment is a computer in which a program for causing a computer to execute a predetermined means (or for causing a computer to function as a predetermined means or for causing a computer to realize a predetermined function) is recorded. It can also be implemented as a readable recording medium.

The present invention is not limited to the above-described embodiments, but can be implemented with various modifications within the technical scope thereof.

[0077]

According to the present invention, prosodic parameters generated by using a plurality of prosodic control dictionaries are interpolated with arbitrary weights to generate prosodic parameters, and speech synthesis is performed using the prosodic parameters. Thus, it is possible to generate synthetic speech having various prosodic features.

Further, according to the present invention, the calculation amount is increased by performing interpolation processing on a plurality of prosody control dictionaries in advance to generate a prosody control dictionary and performing speech synthesis using the prosody control dictionaries. It is possible to generate synthetic speech having a wide variety of prosodic features.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of a speech synthesizer according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a fundamental frequency pattern generation model.

FIG. 3 is a diagram for explaining an example of a representative pattern.

FIG. 4 is a diagram for explaining an example of a fundamental frequency pattern.

FIG. 5 is a diagram for explaining an example of an input unit of weight information.

FIG. 6 is a diagram showing an example of weight information designated by the input means of FIG.

FIG. 7 is a diagram for explaining an example of changes in weight information.

FIG. 8 is a diagram showing a configuration example of a speech synthesizer according to a second embodiment of the present invention.

FIG. 9 is a diagram showing an example of an offset generation rule based on quantification type I.

FIG. 10 is a diagram showing a configuration example of a conventional general speech synthesizer.

FIG. 11 is a diagram showing a configuration example of a conventional speech synthesizing device for controlling a speech style.

[Explanation of symbols]

10 ... Text analysis section 13 ... Voice synthesizer 20, 30 ... Synthesis parameter generation unit 22 ... Synthesis parameter interpolation unit 24-26, 32 ... Prosodic control dictionary 31 ... Prosody control dictionary interpolator 41 ... Representative pattern selection section 42 ... Offset processing unit 43 ... Pattern connection part 44 ... Offset generation unit 45 ... Representative pattern dictionary 46 ... Representative pattern selection rule 47 ... Offset generation rule 48 ... Basic frequency control dictionary

Continuation of the front page (56) References JP-A-10-11083 (JP, A) JP-A-11-231885 (JP, A) JP-A-2000-47681 (JP, A) JP-A-9-244693 (JP, A) Japanese Patent Laid-Open No. 9-90970 (JP, A) Hideki Sakano et al., Voice morphing by independent operation of envelope and sound source, IEICE Transactions A, February 25, 1998, Vol. J81-A, No. 2, p. 261-268 Takakatsu Yoshimura, et al., Speaker Interpolation in HMM-based Speech Synthesis System, Proceedings of Autumn Meeting of Acoustical Society of Japan, 1997, September 1997, 1-P-17, p. 337- 338 (58) Fields surveyed (Int.Cl. ⁷ , DB name) G10L 13/06 G10L 13/08 JISST file (JOIS)

Claims (7)

(57) [Claims] 1. At least according to the input language information.
A first prosody parameter is generated using each of the three prosody control dictionaries, and interpolation processing is performed between the first prosody parameters according to the weight information specified for each prosody control dictionary to generate a second prosody parameter. Is generated , synthetic speech is generated according to the second prosody parameter , and the weight information changes in the sentence . 2. The prosodic control dictionary has a typical fundamental frequency.
Characterized by including a representative pattern that represents a change pattern
The voice synthesis method according to claim 1. 3. A second prosody control dictionary is generated by performing an interpolating process between a plurality of first prosody control dictionaries, and the second prosody control dictionary is used to generate a prosody parameter according to the input linguistic information. In the first prosodic control dictionary , wherein the first and second prosodic control dictionaries have a typical fundamental frequency.
Characterized by including a representative pattern that represents a change pattern
Voice synthesis method. 4. At least according to the input language information.
A means for generating a first prosody parameter using each of the three prosody control dictionaries, and an interpolation process between the first prosody parameters according to the weight information designated for each prosody control dictionary Means for generating a prosody parameter, means for generating a synthetic voice according to the second prosody parameter, and the weight information
A voice synthesizer , comprising: means for changing information in a sentence . 5. Means for performing a interpolation process between a plurality of first prosody control dictionaries to generate a second prosody control dictionaries, and using the second prosody control dictionaries according to input linguistic information. The first and second prosody control dictionaries are provided with a typical basic prosody, and means for generating a prosody parameter and means for generating a synthetic voice according to the prosody parameter.
A speech synthesizer comprising a representative pattern representing a frequency change pattern . 6. At least according to the input language information.
A first prosody parameter is generated using each of the three prosody control dictionaries, and interpolation processing is performed between the first prosody parameters according to the weight information specified for each prosody control dictionary to generate a second prosody parameter. to generate, the second to produce a synthesized speech in accordance with the prosodic parameters, computer-readable recording medium recording the order of the program is changed the weight information in the text. 7. A representative pattern representing a typical fundamental frequency change pattern by performing interpolation processing among a plurality of first prosody control dictionaries.
A second prosody control dictionary including turns is generated, a prosody parameter is generated using the second prosody control dictionary according to the input language information, and a program for generating synthetic speech according to the prosody parameter is recorded. Computer-readable recording medium.