JPH09179576A - Voice synthesizing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To optionally vary the voice quantity of a synthesized voice for text voice synthesis. SOLUTION: When a synthesized voice signal 107 is generated by connecting information 106 selected out of plural pieces of information in voice synthesis units stored in a synthesis unit set storage part 13 by a voice signal generation part 12, a voice signal 201 is inputted from a voice input part 21, pieces of information 204 in voice synthesis units are generated by a voice analysis part 23 according to the voice signal and stored in the synthesis unit set storage part 13, and the synthesized voice signal 107 is generated by using them.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech synthesis method for generating a synthesized speech signal from information such as a phoneme symbol string, pitch and phoneme duration in a text-to-speech system.

[0002]

2. Description of the Related Art Artificially producing a voice signal from an arbitrary sentence (text) is called text-to-speech synthesis. Normally, the text-to-speech synthesis system is composed of three components, a language processing unit 10, a phoneme processing unit 11, and a speech signal generation unit 12, as shown in FIG.

The input text 101 is first subjected to morphological analysis and syntactic analysis in the language processing unit 10,
Next, accent processing and intonation processing are performed in the prosody processing unit 11, and information such as the phoneme symbol string 103, the phoneme duration 104, and the pitch pattern 105 is output. Finally, in the voice signal generation unit 12, the characteristic parameters of a small basic unit (speech synthesis unit) such as syllables, phonemes, and 1-pitch sections stored in the synthesis unit set storage unit 13 are set to the phonological symbol sequence 103, phonological continuation. Selection is made according to information such as the time length 104 and the pitch pattern 105, and the synthesized voice signal 107 is generated by connecting these by controlling the pitch and duration.

The voice quality (individuality) of the synthesized speech thus generated is mainly determined by the characteristic parameter which is a speech synthesis unit. Therefore, when it is necessary to synthesize a plurality of voices having different voice qualities in text-to-speech synthesis, a plurality of synthesis unit set storage units 13 and 14 corresponding to the respective voice qualities are prepared as shown in FIG. By switching and using, the synthesized voice signal 107 of different voice quality is generated.

On the other hand, as a method of synthesizing voices of various voice qualities without storing a plurality of synthesis unit sets, a method of changing a pitch pattern, a method of changing a sound quality by filtering a synthesized voice signal, etc. There is.

The former method of changing the pitch pattern is to change the pitch of the voice by inputting the pitch average value information 109 into the prosody processing section 15 and changing the average value of the pitch pattern as shown in FIG. By inputting the intonation level information 110 and changing the width of the pitch change, the intonation intensity is changed. The latter method using a filter changes the sound quality of synthesized speech by emphasizing the low and high frequencies of the synthesized speech signal 107 by a filter unit 16 whose characteristics can be controlled according to filter characteristic information 111 as shown in FIG. Is.

[0007]

However, the conventional speech synthesis method as described above has a problem in that it is not possible to generate synthesized speech having an arbitrary voice quality. That is, in the method of storing a plurality of synthesis unit sets, it is possible to synthesize a voice of a plurality of types of voice quality by increasing the number of synthesis unit sets, but it is practical to synthesize a voice of any voice quality. However, there is a problem in that it is impossible to store a large number of composition unit sets, and the required memory capacity increases and the cost increases.

Further, in the method of changing the pitch pattern, even if the intonation and accent are changed, it is essentially impossible to change the voice quality of the synthesized voice, that is, the individuality. When the voice quality of the synthesized voice is changed by the filter, the voice can be changed to be brighter / darker, but the individuality cannot be changed. This is because the personality information exists not only in the amplitude of the formant but also in the center frequency and width of the formant.

The present invention has been made to solve the above problems, and an object of the present invention is to provide a speech synthesis method capable of arbitrarily changing the voice quality of synthesized speech in text speech synthesis.

[0010]

In order to solve the above problems, the present invention provides a voice synthesizing method for generating a synthetic voice signal by connecting information selected from information of a plurality of voice synthesizing units. According to at least one of the voice quality information and the voice quality information, information of a plurality of voice synthesis units necessary for generating a synthetic voice signal is generated or processed, and specifically, the following features are adopted.

(1) Input a voice signal and generate information of a plurality of voice synthesis units according to the voice signal. If the information of the voice synthesis unit generated according to the voice signal is stored and the synthesized voice signal is generated using this, the voice quality of the synthesized voice is the same as the voice quality of the speaker who inputs the voice signal. Therefore, the voice quality of the synthesized voice changes when the speaker changes.

(2) Input a voice signal and process information of a plurality of voice synthesis units stored in advance according to the voice signal. Thus, if the information of the existing voice synthesis unit is processed according to the input voice signal and the synthesized voice signal is generated using this, the voice quality of the synthesized voice becomes the same as that of the speaker who inputs the voice signal. , The voice quality of synthetic speech changes depending on the speaker.

(3) Input voice quality information, and process information of a plurality of voice synthesis units stored in advance according to this voice quality information. Here, the voice quality information is information that specifies the voice quality of the synthesized voice. If the information of the existing voice synthesis unit is processed according to this voice quality information and a synthesized voice signal is generated using this, the voice quality of the synthesized voice changes according to the voice quality information.

(4) Input a voice signal and voice quality information,
Information of a plurality of voice synthesis units is generated from a voice signal, and information of a plurality of voice synthesis units is processed according to voice quality information. In this way, after the information of the voice synthesis unit is generated according to the voice signal, the processing according to the voice quality information is performed, and if the synthesized voice signal is generated using this, the voice quality of the synthesized voice is the input voice signal and voice quality information. According to both, it will change more variably.

(5) Input a voice signal and voice quality information,
Information on a plurality of voice synthesis units stored in advance is processed according to these voice signals and voice quality information. In this case as well, the voice quality of the synthesized voice is variously changed according to both the input voice signal and the voice quality information.

(6) A plurality of voice quality information is stored, and the information of the voice synthesis unit stored in advance is processed according to the voice quality information. In this way, with a smaller memory capacity than when storing the information of the synthesis unit corresponding to a plurality of voice qualities,
It is possible to obtain a synthetic voice having a plurality of voice qualities.

(7) A plurality of voice quality information is stored, and the information of the voice synthesis unit is processed according to the voice quality information generated by interpolating the plurality of voice quality information. By doing this, it is possible to obtain a synthesized voice having a variety of voice qualities such as a mixture of a plurality of voice qualities given by a plurality of voice quality information.

In the present invention, when processing the information of a plurality of voice synthesis units, all the voice synthesis units may be processed, but only the voice synthesis unit including a vowel mainly containing voice quality information is processed. It may be performed. Further, the processing of the information of the voice synthesis unit is performed on the spectrum information by, for example, at least one of amplitude mapping and frequency mapping.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION

(First Embodiment) FIG. 1 shows the configuration of a voice synthesizing apparatus according to a first embodiment to which the first voice synthesizing method of the present invention is applied. This speech synthesizer includes a speech input section 21, a speech recognition section in addition to the conventional speech synthesizer shown in FIG. 12 having a language processing section 10, a phoneme processing section 11, a speech signal generation section 12, and a synthesis unit set storage section 13. 22 and a voice analysis unit 23 are added.

The basic operation of this speech synthesizer is the same as the conventional one. That is, the input text 101 is first subjected to morphological analysis and syntactic analysis in the language processing unit 10, and then subjected to accent and intonation processing in the prosody processing unit 11 based on the analysis result 102 from the language processing unit 10. Then, information such as the phoneme symbol string 103, the phoneme duration 104 and the pitch pattern 105 is output. Then, in the speech signal generation unit 12, the characteristic parameters of a small basic unit (speech synthesis unit) such as a syllable, a phoneme, and one pitch section stored in the synthesis unit set storage unit 13 are the phonological symbol sequence 103, the phonological duration. The length 104 and the pitch pattern 105 are selected according to the information, and these are connected after the pitch and the duration are controlled, whereby the synthesized voice signal 107 is selected.
Is generated.

Next, the characteristic structure of the present embodiment will be described. First, the voice input unit 21 converts a voice generated by a speaker into an electric signal by a microphone, and further performs an appropriate process such as amplification to output the voice. It is configured to be output as the signal 201, or the audio signal supplied from an external device is input as it is or after being subjected to appropriate processing.

The voice recognition unit 22 associates the voice signal 201 input from the voice input unit 21 with the phoneme symbol, and at the same time, corresponds to the phoneme indicated by the phoneme symbol corresponding to the input voice signal 201. The section is determined and the phoneme symbol 202 and the phoneme boundary time information 203 are output.

The voice analysis unit 23 analyzes the voice signal 201 in the section serving as the voice synthesis unit from the voice input unit 21 according to the phoneme boundary time information 203 and the phoneme symbol 202 input from the voice recognition unit 22, and outputs the voice. The characteristic parameter 204 of the composition unit is output.

Then, the characteristic parameter 204 output from the voice analysis unit 23 is stored in the synthesis unit set storage unit 13 as information of the voice synthesis unit. An example of the characteristic parameter is a pair of LPC coefficient and residual waveform. In addition to this, it is also possible to use various parameters such as formants and cepstrums.

In the speech synthesizing apparatus having the above-described configuration, in the present embodiment, the speech input unit 21 inputs the speech signal 201 including all speech synthesizing units in advance, and the characteristic parameters of all speech synthesizing units are used as the synthesizing units. Set storage unit 1
After storing in 3, the text composition processing is performed. In this case, the voice quality of the synthesized voice obtained by reproducing the synthesized voice signal 107 by the speaker is the same as the voice quality input by the voice input unit 21 to generate the synthesis unit set. If it is desired to change the voice quality of the synthesized voice, the voice synthesis unit set may be regenerated using the voice of the speaker having the target voice quality.

As described above, according to this embodiment, in the text-to-speech synthesis, it is possible to generate a synthesized voice having an arbitrary voice quality that matches the voice quality of the speaker who inputs the voice signal by the voice input unit 21.

(Second Embodiment) FIG. 2 shows the configuration of a voice synthesizing apparatus according to a second embodiment to which the voice synthesizing method of the present invention is applied. In FIG. 2, elements given the same reference numerals as those in FIG. 1 have the same functions as those in FIG.

In this embodiment, a phonological unit presenting unit 28 for presenting phonological symbols to be uttered by a speaker is provided, and this phonological unit presenting unit 28 is provided.
The voice signal input from the voice input unit 21 is input to the voice analysis unit 23 when the speaker utters the phoneme indicated by the phoneme symbol presented by.
The phoneme presentation unit 28 may be, for example, one that visually displays a phoneme symbol on a display, or that that is presented by voice through a speaker.

According to this embodiment, the same effect as that of the first embodiment can be obtained, and since the voice analysis unit 23 knows the phoneme of the voice signal input from the voice input unit 21, the voice shown in FIG. The recognition unit 22 may be omitted. Considering that the recognition rate of the voice recognition device is not sufficiently high under the present circumstances, the present embodiment is effective in increasing the reliability of the synthesis unit set stored in the synthesis unit set storage unit 13.

(Third Embodiment) FIG. 3 shows the configuration of a speech synthesizer according to a third embodiment to which the speech synthesis method of the present invention is applied. In FIG. 3, elements given the same reference numerals as those in FIG. 1 have the same functions as those in FIG.

In the present embodiment, in addition to the configuration shown in FIG. 1, a spectrum comparison section 24, a synthesis unit information conversion section 25 and another synthesis unit set storage section 27 are provided. The first synthesis unit set storage unit 26 stores a predetermined feature parameter of a voice synthesis unit, and the feature parameter obtained by converting the voice quality by processing this feature parameter is stored in the second synthesis unit set storage unit. Stored in 27.

The spectrum comparing section 24 is provided in the voice input section 21.
Of the spectrum parameter of the spectrum parameter of the feature parameter 204 obtained by analyzing the voice signal 201 from the voice analysis unit 23 and the spectrum parameter of the feature parameter 205 stored in the first synthesis unit set storage unit 26. The power spectrum is compared and the conversion parameter 206 is output. The composition unit information conversion unit 25
By converting the characteristic parameter 207 based on the conversion parameter 206 from the spectrum comparison unit 24, the characteristic parameter 208 of the voice synthesis unit whose voice quality has been converted is output and stored in the second synthesis unit set storage unit 27. .

The processing of the synthesis unit information conversion section 25 can be realized by, for example, mapping frequency and amplitude with respect to spectrum parameters. In this case, the spectrum comparison unit 24 outputs the mapping function of frequency or amplitude as the conversion parameter 206. If f is the frequency,
Power spectrum s (f) in the characteristic parameter 205
From the power spectrum s ′ in the characteristic parameter 204
The conversion into (f) is expressed by the following equation.

S ′ (f) = m (s (ω (f))) (1) where ω () is a function representing frequency mapping and m () is a function representing amplitude mapping.

For example, s (f) and s' (f) are shown in FIG.
It is assumed that they are represented by (a) and (b), respectively. Where F
₁ , F ₂ , F ′ ₁ and F ′ ₂ are formant frequencies,
P ₁ , P ₂ , P ′ ₁ and P ′ ₂ are the powers of the corresponding formants. In this case, the frequency mapping function ω () corresponding to the conversion parameter 206 is a function for mapping F ₁ to F ′ ₁ and F ₂ to F ′ ₂ , respectively, and the amplitude mapping function m () is It is a function that maps P ₁ to P ′ ₁ and P ₂ to P ′ ₂ . An example of these functions ω () and m () is shown in FIG.
(A) and (b) respectively show.

In the present embodiment, regarding the conversion of the characteristic parameters in the speech synthesis unit information conversion section 25, for example, the corresponding conversion parameters are obtained for all the speech synthesis units, and different conversion may be performed for each characteristic parameter. It is also possible to classify the speech synthesis units into several classes and obtain common conversion parameters for one class for conversion. For example, a plurality of power spectra s _i (f), (i = 1, 2, 3, ..., N)
Corresponding power spectra s ′ _i (f), (i = 1,
2, 3, ..., N) common mapping function ω
() And m () are obtained by minimizing a conversion error function E (ω, m) represented by the following equation using a known optimization method.

[0037]

[Equation 1]

However, fs is a sampling frequency. Since voice quality information is mainly contained in vowels, it is possible to convert the voice quality by converting feature parameters only for a voice synthesis unit including a vowel and using the other synthesis units without conversion. It is possible. By doing so, the amount of processing in the synthesis unit information conversion unit 25 can be reduced.

Further, the synthesis unit information conversion section 25 can also perform conversion as in the following equation. s' (f) = g (f) s (ω (f)) (3) This is an example in which the amplitude conversion is realized by the product of the function g (f).

In the present embodiment, the voice synthesis unit set stored in advance in the first synthesis unit set storage unit 26,
Any of the voice synthesis unit sets generated by converting the voice quality and stored in the second synthesis unit set storage unit 27 can be used as the feature parameter 106 of the voice synthesis unit in the voice signal generation unit 12. Therefore, according to the voice quality selection information 108 that specifies the voice quality of the synthesized voice, the first
By synthesizing and using the synthesizing unit set storage unit 26 and the second synthesizing unit set storage unit 27, it is possible to perform speech synthesis with various voice qualities.

Further, as in the second embodiment, the present embodiment can be modified to a structure in which a phoneme presentation unit is provided and a voice recognition unit is not used. (Fourth Embodiment) FIG. 6 shows the configuration of a voice synthesizing apparatus according to a fourth embodiment to which the voice synthesizing method of the present invention is applied. In FIG. 6, elements given the same reference numerals as those in FIG. 3 have the same functions as those in FIG.

In the third embodiment, a voice synthesis unit set in which the voice quality is converted is created and stored in advance. In the present embodiment, the voice quality information is stored in the form of a conversion parameter, and each time voice synthesis is performed, the voice quality information is stored. The characteristic parameters are converted and supplied to the audio signal generator.

That is, in the present embodiment, the second unit for storing the voice synthesis unit set obtained by converting the voice quality in FIG.
The synthesis unit set storage unit 27 is removed and a conversion parameter storage unit 32 is provided instead. The conversion parameter storage unit 32 stores the conversion parameter 308 obtained by the spectrum comparison unit 24, and the conversion parameter 206 corresponding to the voice synthesis unit used when performing voice synthesis.
Is output. The synthesis unit information conversion unit 25 converts the characteristic parameter 207 based on the conversion parameter 206 and outputs the characteristic parameter 208, which is supplied to the audio signal generation unit 12.

According to this embodiment, the synthesis unit set storage unit 2 for storing the voice synthesis unit whose voice quality has been converted in FIG.
Since the memory amount of the conversion parameter storage unit 32 is much smaller than the memory amount of 7, it is possible to reduce the memory amount required for the speech synthesizer as compared with the third embodiment.

(Fifth Embodiment) FIG. 7 shows the arrangement of a speech synthesis apparatus according to the fourth embodiment to which the speech synthesis method of the present invention is applied. In FIG. 7, elements given the same reference numerals as those in FIG. 3 have the same functions as those in FIG.

In this embodiment, a voice quality input unit 31, a conversion parameter generation unit 30, a conversion parameter storage unit 36, and a conversion parameter input / output unit 37 are provided. The voice quality input unit 31 outputs voice quality information 301 that specifies how to convert the quality of the synthesized voice on a scale that the user can understand. The scales that the user can understand include, for example, “bright voice / dark voice”, “thick voice / thin voice”, “child voice / adult voice”, and other various scales. There is.
The conversion parameter generation unit 30 generates a conversion parameter 302 so as to perform voice quality conversion according to the voice quality information 301.

The conversion parameter storage unit 36 stores the conversion parameter 302 and the conversion parameter 310 input from the conversion parameter input / output unit 37, and the voice quality selection information 3
The conversion parameter 309 is output according to 06. The conversion parameter input / output unit 37 outputs the stored conversion parameter 311 to the outside or inputs the conversion parameter 310 from the outside, and is, for example, a storage device such as a magnetic storage device or a CD-ROM, or a network. It is realized by the interface with.

As described above, according to the present embodiment, the conversion parameter 302 generated by the conversion parameter generation unit 30 based on the voice quality information 301 input by the voice quality input unit 31 in accordance with the instruction of various methods is used. It is possible to synthesize a voice having the voice quality of, and the generated conversion parameter 302 is stored in the conversion parameter storage unit 36.
With the configuration that allows input and output of conversion parameters with the outside, it is also possible to provide the generated conversion parameters to other users and, conversely, perform voice quality conversion using the conversion parameters generated by other users. Is.

(Sixth Embodiment) FIG. 8 shows the arrangement of a speech synthesis apparatus according to the sixth embodiment to which the speech synthesis method of the present invention is applied. In FIG. 8, elements given the same reference numerals as those in FIGS. 1 and 7 have the same functions as those in FIGS. 1 and 7, and a description thereof will be omitted.

In the present embodiment, a synthesized voice segment set having a desired voice quality is generated by designating how to change the voice quality of an input voice. That is, the feature parameter 204 obtained by analyzing the voice signal 101 input from the voice input unit 21 by the voice analysis unit 23 is input to the synthesis unit information conversion unit 25, and the voice quality information 301 input from the voice quality input unit 31 is input. Based on the conversion parameter generation unit 30, the synthesis unit information conversion unit 25 generates a voice synthesis unit characteristic parameter 208 in which the voice quality is converted according to the conversion parameter 302 generated by the conversion parameter generation unit 30.
To memorize.

As described above, according to this embodiment, after the characteristic parameter 204 of the voice synthesis unit is generated according to the voice signal 201, the processing is performed according to the voice quality information 301, and the synthesized voice signal 107 is generated using this. As a result, the voice quality of the synthesized voice can be variously changed according to both the input voice signal 201 and voice quality information 301.

(Seventh Embodiment) FIG. 9 shows the configuration of a speech synthesizing apparatus according to the seventh embodiment to which the speech synthesizing method of the present invention is applied. In FIG. 9, elements designated by the same reference numerals as those in FIGS. 2 and 7 have the same functions as those in FIGS. 2 and 7, and a description thereof will be omitted.

In this embodiment, the characteristic parameter 207 of the voice synthesis unit stored in the first synthesis unit set storage unit 26 in advance is close to the voice quality of the voice signal 201 input by the synthesis unit information conversion unit 25. After converting to
The converted characteristic parameter 303 is subjected to the voice quality conversion specified by the voice quality information 301 from the voice quality input unit 31 in the other synthesis unit information conversion unit 33 to obtain a synthesized voice segment set of a desired voice quality. The characteristic parameter 208 is generated, and the second synthesis unit set storage unit 2 is generated.
It is designed to be stored in 7. Therefore, according to the present embodiment, similarly to the sixth embodiment, the voice quality of the synthesized voice can be variously changed according to both the input voice signal 201 and voice quality information 301.

(Eighth Embodiment) FIG. 10 shows the configuration of a speech synthesis apparatus according to an eighth embodiment to which the speech synthesis method of the present invention is applied. In FIG. 10, elements given the same reference numerals as those in FIG. 3 have the same functions as those in FIG.

The conversion parameter storage unit 36 stores conversion parameters corresponding to several voice qualities in advance, and the conversion parameter 302 selected according to the voice quality selection information 306.
Is output. According to the present embodiment, similar voice quality conversion can be realized with a smaller memory amount than that in which voice synthesis unit sets corresponding to a plurality of voice qualities are stored in advance. Further, as a modification of the present embodiment, it is possible to store sample voices corresponding to voice qualities, present the voices, and select the voice qualities.

(Ninth Embodiment) FIG. 11 shows the configuration of a speech synthesizing apparatus according to the ninth embodiment to which the speech synthesizing method of the present invention is applied. 11, elements given the same reference numerals as those in FIG. 10 have the same functions as those in FIG.

The conversion parameter storage unit 34 selects and outputs a plurality of conversion parameters 312 according to the voice quality selection information 305. The conversion parameter interpolation unit 35 interpolates the plurality of conversion parameters 312 according to the weighting coefficient 307 and outputs the conversion parameters 302. According to this embodiment, it is possible to synthesize a voice having a voice quality in which some voice qualities whose conversion parameters are stored in advance are mixed at an arbitrary ratio.

[0058]

As described above, according to the voice synthesizing method of the present invention, a voice signal or voice quality information is input and a voice having a voice quality similar to that of the voice signal or a voice quality specified by the voice quality information is obtained. It is possible to synthesize a voice, and the voice quality of the synthesized voice can be arbitrarily changed in the text voice synthesis.

[Brief description of the drawings]

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

FIG. 2 is a block diagram showing a configuration of a speech synthesizer according to a second embodiment of the present invention.

FIG. 3 is a block diagram showing a configuration of a speech synthesizer according to a third embodiment of the present invention.

FIG. 4 is a diagram showing an example of a characteristic parameter obtained by analyzing an input voice in the same embodiment and a power spectrum of a spectrum in the stored characteristic parameters.

FIG. 5 is a diagram showing an example of frequency and amplitude mapping functions corresponding to conversion parameters in the same embodiment.

FIG. 6 is a block diagram showing the configuration of a speech synthesizer according to a fourth embodiment of the present invention.

FIG. 7 is a block diagram of a block diagram showing the configuration of a speech synthesizer according to a fifth embodiment of the present invention.

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

FIG. 9 is a block diagram of a block diagram showing the configuration of a speech synthesizer according to a seventh embodiment of the present invention.

FIG. 10 is a block diagram of a block diagram showing the configuration of a speech synthesizer according to an eighth embodiment of the present invention.

FIG. 11 is a block diagram of a block diagram showing the configuration of a speech synthesizer according to a ninth embodiment of the present invention.

FIG. 12 is a block diagram showing the configuration of a conventional speech synthesizer.

FIG. 13 is a block diagram showing the configuration of a conventional speech synthesizer that changes the voice quality of synthesized speech by storing a plurality of speech synthesis unit sets.

FIG. 14 is a block diagram showing the configuration of a conventional speech synthesizer that changes the voice quality of synthesized speech by changing the pitch pattern.

FIG. 15 is a block diagram showing a configuration of a speech synthesizer that changes a voice quality by a conventional filter.

[Explanation of symbols]

10 ... Language processing section 11 ... Prosody processing section 12 ... Speech signal generation section 13, 14 ... Synthesis unit set storage section 15 ... Prosody processing section 16 ... Filter section 21 ... Speech input section 22 ... Speech recognition section 23 ... Speech analysis section 24 ... spectrum comparison unit 25 ... synthesis unit information conversion unit 26, 27 ... synthesis unit set storage unit 28 ... phoneme presentation unit 30 ... conversion parameter generation unit 31 ... speech input unit 32 ... conversion parameter storage unit 33 ... synthesis unit information conversion unit 34 ... conversion parameter storage unit 35 ... conversion parameter interpolation unit 36 ... conversion parameter storage unit 37 ... conversion parameter input / output unit 101 ... input text 102 ... analysis data 103 ... phoneme symbol string 104 ... phoneme duration 105 ... pitch pattern 106 ... speech Characteristic parameter 107 of synthesis unit ... Synthetic voice signal 108 ... Voice quality conversion information 201 ... Voice signal 202 ... Phonological symbol 203 ... Phonological boundary time information 204, 205 ... Feature parameter 206 ... Transformation parameter 207, 208 ... Feature parameter 301 ... Voice quality information 302 ... Transformation parameter 303 ... Feature parameter 305, 306 ... Voice quality selection information 307 ... Weighting coefficient 308 , 309, 310, 311, 312 ... Conversion parameters

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Masahiro Oshikiri No. 1 Komukai Toshiba-cho, Sachi-ku, Kawasaki City, Kanagawa Prefecture, Corporate Research & Development Center, Toshiba Corp. Town No. 1 Incorporated Toshiba Corporation R & D Center (72) Inventor Akinobu Yamashita Komukai Toshiba Town No. 1 Komukai Toshiba Town Kanagawa Prefecture Kanagawa Prefecture

Claims (7)

[Claims] 1. A voice synthesizing method for generating a synthesized voice signal by connecting information selected from information of a plurality of voice synthesizing units, wherein a voice signal is input, and information of a plurality of voice synthesizing units is input in accordance with the voice signal. A method for synthesizing speech, comprising: 2. A voice synthesizing method for generating a synthesized voice signal by connecting information selected from information of a plurality of voice synthesizing units, wherein a voice signal is input, and a plurality of voice synthesizing stored in advance according to the voice signal. A speech synthesis method characterized by processing unit information. 3. A voice synthesizing method for generating a synthesized voice signal by connecting information selected from information of a plurality of voice synthesizing units, wherein voice quality information is input and a plurality of voice synthesizing stored in advance according to the voice quality information. A speech synthesis method characterized by processing unit information. 4. A voice synthesizing method for generating a synthesized voice signal by connecting information selected from information of a plurality of voice synthesizing units, wherein a voice signal and voice quality information are inputted, and a plurality of voice synthesizing is performed by said voice signal. A method for synthesizing voice, comprising generating unit information and processing information on a plurality of voice synthesizing units according to the voice quality information. 5. A voice synthesizing method for generating a synthetic voice signal by connecting information selected from information of a plurality of voice synthesizing units, wherein voice signal and voice quality information are input, and according to these voice signal and voice quality information. A voice synthesizing method characterized by processing information of a plurality of voice synthesizing units stored in advance. 6. A voice synthesizing method for generating a synthesized voice signal by connecting information selected from information of a plurality of voice synthesizing units, wherein a plurality of voice quality information is stored, and the voice synthesizing is stored in advance in accordance with the voice quality information. A speech synthesis method characterized by processing unit information. 7. A voice synthesizing method for generating a synthesized voice signal by connecting information selected from a plurality of stored voice synthesizing unit information, storing a plurality of voice quality information and interpolating the plurality of voice quality information. A method of synthesizing voice, characterized by processing information of a voice synthesis unit according to voice quality information generated by