JP2012073338A - Speech synthesizer and speech synthesis method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a speech synthesizer capable of reducing time required for generating synthesized speech even when the synthesize speech is generated by using any of plural synthesis conditions different from one another for a predetermined original.SOLUTION: A speech synthesizer 1 includes: a matching determination unit 10 for determining whether a set of input text information or input phonetic information and an input synthesis condition matches any of sets of at least one piece of registered text information or registered phonetic information and a registered synthesis condition; and an automatic update unit 17 for, when the set of the input text information and the like does not match any of the sets of the registered text information and the like, generating a set of derivative text information and derivative synthesis condition by correcting the input text information or the input synthesis condition, and making a storage unit 3 store a set of a set of the derivative text information and the derivative synthesis condition and derivative intermediate information generated during generation of a synthesized speech signal by a speech synthesis unit 11 based on the derivative text information and the derivative synthesis condition, as a set of a set of the registered text information and the registered synthesis condition and registered intermediate information.

Description

  The present invention relates to a speech synthesizer and a speech synthesis method for synthesizing speech signals, for example.

  In recent years, speech synthesis technology for automatically synthesizing speech has been developed. Since speech synthesis technology has the advantage that it can create desired speech in a short time, such speech synthesis technology has been adopted in applications that have used pre-recorded speech by professional narrators. There are also things. In particular, in an application in which information provided at a short time interval is updated, such as a guidance broadcast in a commercial facility, a highway radio, a highway telephone, or a weather forecast broadcast, the speech synthesis technology having the above-described advantages is useful. However, in such an application, for a device that creates synthesized speech using speech synthesis technology, the quality of the synthesized speech is high so that the content of the message is easy for the viewer to understand, and a large amount in a short time. It is required to be able to create synthesized speech.

  Therefore, in an example of known technology, the speech synthesizer selectively selects any one of the original text information, the phoneme string information, and the speech waveform information according to the reference frequency of the voice output request for the original text information with respect to the once created voice. accumulate. This speech synthesizer shortens the time required for speech synthesis by using that information for speech waveform synthesis if there is already accumulated information about the speech to be output (for example, Patent Documents). 1).

Japanese Patent Laid-Open No. 5-19790

In the above-described conventional technology, the speech synthesizer stores only intermediate information such as one type of phoneme information created based on a specific synthesis condition such as a specific speech speed or pitch for one original sentence. However, the speech synthesizer may create output speech with different speech speeds or different pitches (ie, voice pitch) for one original sentence. For example, with regard to narration used in commercial facility broadcasting, it is preferable that the speech speed in an emergency (for example, when a fire occurs) be faster than the normal speech speed. However, if the speech synthesizer uses intermediate information such as phoneme information created under a specific synthesis condition for a certain original sentence to create a voice output under other synthesis conditions for that original sentence, a newly generated The prosody of synthesized speech is unnatural. As a result, the quality of the newly generated synthesized speech deteriorates. In addition, the quality of the synthesized speech that has been corrected by directly converting the synthesized speech that has been created is lower than the quality of the original synthesized speech.
In addition, when it is required to create a synthesized speech under a synthesis condition that is different from the synthesis conditions used to create the stored intermediate information, the speech synthesizer does not use the intermediate information and the original text If the synthesized speech is created again from the above, the time required for speech synthesis cannot be shortened.

  Accordingly, the present specification provides a speech synthesizer and a speech synthesis method that can reduce the time required to create a synthesized speech even when a synthesized speech is created with any of a plurality of different synthesis conditions for a predetermined original text. The purpose is to do.

  According to one embodiment, a speech synthesizer is provided. The speech synthesizer includes an input unit that obtains input text information including an original sentence that is a source of a synthesized speech signal, an input synthesis condition for creating a synthesized speech signal, a set of registered text information and a registered synthesis condition, A storage unit for storing at least one set of registration intermediate information including registration phonetic information, which is generated in the middle of creating a synthesized speech signal corresponding to the set of registered text information and registration synthesis condition; A coincidence determination unit that determines whether a pair of input phonetic information represented by text information or input text information and an input synthesis condition matches any of registered text information or a set of registered phonetic information and registered synthesis conditions And the combination of the input text information or the input phonetic information and the input synthesis condition matches the registered text information or the set of the registered phonetic information and the registration synthesis condition. Information or registered phonetic information and registered synthesis information are used to create a synthesized speech signal, while input text information or input phonetic information and input synthesis conditions are registered text information. Or, if it does not match any of the set of registered phonetic information and registered synthesis condition, a speech synthesis unit that creates a synthesized speech signal based on the input text information and the input synthesis condition, and input text information or input phonetic information and input If the combination of the composition condition does not match any of the registered text information or registered phonetic information and the registered composition condition, the combination of the derived text information and the derived composition condition is changed by modifying the input text information or the input composition condition. The speech synthesizer creates a synthesized speech signal based on the combination of the derived text information and derived synthesis condition, and the derived text information and derived synthesis condition. An automatic update unit for storing a set of the derived intermediate information created by executing up to the intermediate stage in the storage unit as one of a set of registered text information, a set of registered composition conditions, and registered intermediate information. Have.

  According to another embodiment, a speech synthesis method is provided. This speech synthesis method acquires input text information including an original text that is a source of a synthesized speech signal and input synthesis conditions for creating a synthesized speech signal, and is an input table represented by the input text information or the input text information. It is determined whether the set of the sound information and the input synthesis condition matches at least one of the registered text information or the set of the registered phonetic information and the registration synthesis condition stored in the storage unit, and the input text information Alternatively, when the set of the input phonetic information and the input synthesis condition matches either of the registered text information or the set of the registered phonetic information and the registration synthesis condition, it is stored in the storage unit, and the registered text information or the registered phonetic information A synthesized speech signal is created using registered intermediate information generated in the middle of creating a synthesized speech signal corresponding to a set of registered synthesis conditions. Creates a synthesized speech signal based on the input text information and input synthesis conditions if the set of input phonetic information and input synthesis conditions does not match any of the registered text information or registered phonetic information and registration synthesis conditions If the combination of the input text information or the input phonetic information and the input synthesis condition does not match any of the registered text information or the registered phonetic information and the registration synthesis condition, the input text information or the input synthesis condition is corrected. To create a set of derived text information and derived synthesis conditions, and a set of derived text information and derived synthesis conditions, and a derivation created during the creation of a synthesized speech signal based on the derived text information and derived synthesis conditions Storing the set of intermediate information in the storage unit as one of a set of registered text information, a set of registered composition conditions, and registered intermediate information.

The objects and advantages of the invention will be realized and attained by means of the elements and combinations particularly pointed out in the appended claims.
It should be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not restrictive of the invention as claimed.

  The speech synthesizer and the speech synthesis method disclosed in this specification reduce the time required for creating a synthesized speech even when a synthesized speech is created with any of a plurality of different synthesis conditions for a predetermined original text. it can.

It is a schematic block diagram of the speech synthesizer by 1st Embodiment. It is a schematic block diagram of the process part which the speech synthesizer by 1st Embodiment has. It is a figure which shows an example of an intermediate information table. It is a figure which shows another example of an intermediate information table. It is a figure which shows the operation | movement flowchart of the speech synthesis process by 1st Embodiment. It is a schematic block diagram of the process part of the speech synthesizer by 2nd Embodiment. It is a figure which shows an example of the intermediate information table by 2nd Embodiment. It is an operation | movement flowchart of the survival determination process of registration intermediate information. It is a schematic block diagram of the process part of the speech synthesizer by 3rd Embodiment. It is a figure which shows another example of the intermediate information table by a modification.

Hereinafter, speech synthesis apparatuses according to various embodiments will be described with reference to the drawings.
This speech synthesizer creates and stores intermediate information even under other synthesis conditions when creating a new synthesized speech for a given original text under specific synthesis conditions. The time required to create synthesized speech for the predetermined original text is shortened.

  FIG. 1 is a schematic configuration diagram of a speech synthesizer according to one embodiment. In the present embodiment, the speech synthesizer 1 includes an input unit 2, a storage unit 3, a processing unit 4, and an output unit 5.

The input unit 2 acquires text information that is an original text of synthesized speech and synthesis parameters that define speech synthesis conditions such as speech speed, pitch, and voice pitch range. For this purpose, the input unit 2 includes, for example, a keyboard. The input unit 2 includes a pointing device such as a mouse and a display that displays characters or numerical values to be input, which are instructed by the pointing device. Alternatively, the input unit 2 may have a touch panel display.
Furthermore, the input unit 2 may acquire text information and synthesis parameters from another device connected to the speech synthesizer 1 via a communication network. In this case, the input unit 2 includes an interface circuit for connecting the speech synthesizer 1 to a communication network.
Then, the input unit 2 passes the input text information and synthesis parameters to the processing unit 4.

The storage unit 3 includes, for example, at least one of a semiconductor memory circuit, a magnetic storage device, and an optical storage device. The storage unit 3 stores various computer programs used in the processing unit 4 and various data used for speech synthesis processing.
The storage unit 3 stores, for example, a language dictionary, a prosodic model, and a speech waveform dictionary as data used for speech synthesis processing. Further, the storage unit 3 stores an intermediate information table in which intermediate information such as phonetic information or waveform generation information is generated in the middle of creating a synthesized speech signal. Details of the language dictionary, prosodic model, speech waveform dictionary, intermediate information, and intermediate information table will be described later.

The output unit 5 outputs the synthesized voice signal received from the processing unit 4 to the speaker 6. For this purpose, the output unit 5 includes, for example, an audio interface circuit for connecting the speaker 6 to the speech synthesizer 1.
The output unit 5 may output the synthesized speech signal to another device connected to the speech synthesizer 1 via the communication network. In this case, the output unit 5 includes an interface circuit for connecting to the speech synthesizer 1 to the communication network. When the input unit 2 also acquires text information and synthesis parameters via the communication network, the input unit 2 and the output unit 5 may be integrated.

The processing unit 4 includes one or a plurality of processors, a memory circuit, and a peripheral circuit. Then, the processing unit 4 creates a synthesized speech signal based on the original text and synthesis parameters indicated in the input text information, or intermediate information corresponding to the combination of the original text and the synthesis parameters. For this purpose, the processing unit 4 includes a match determination unit 10, a voice synthesis unit 11, a control unit 16, and an automatic update unit 17.
Each of these units included in the processing unit 4 is, for example, a functional module realized by a computer program that operates on a processor included in the processing unit 4. Alternatively, these units included in the processing unit 4 may be mounted on the speech synthesizer 1 as separate circuits. Furthermore, each of these units included in the processing unit 4 may be mounted on the speech synthesizer 1 as one integrated circuit that realizes the function of each unit.

The coincidence determination unit 10 includes text information input via the input unit 2 or a combination of phonetic information represented by the text information and a synthesis parameter, any text information registered in the intermediate information table or It is determined whether or not the set of the phonetic information and the synthesis parameter matches. If the set of input text information or phonetic information and the synthesis parameter matches any text information or phonetic information and the combination of synthesis parameters registered in the intermediate information table, the input text Intermediate information corresponding to a set of information and synthesis parameters is also registered in the intermediate information table. The coincidence determination unit 10 is registered in an intermediate information table that matches a set of input text information or phonetic information and a synthesis parameter in order to use already generated intermediate information for generating a synthesized speech signal. The identification number of the combination of the text information or the phonetic information and the synthesis parameter is passed to the control unit 16.
Note that text information, synthesis parameters, and intermediate information registered in the intermediate information table are hereinafter referred to as registered text information, registered synthesis parameters, and registered intermediate information for convenience.

On the other hand, if the set of input text information and synthesis parameter does not match any of the registered text information and registered synthesis parameter set, the speech synthesizer 1 synthesizes synthesized speech based on the input text information and synthesis parameter set. Create a signal. Accordingly, the coincidence determination unit 10 passes the set of the input text information and the synthesis parameter to the speech synthesis unit 11.
Furthermore, text information including a text that is partly different from the original text of the input text information may be input later, or a composite parameter in which at least one of the input composite parameters is different may be input later. Therefore, in such a case, the match determination unit 10 passes the input text information and synthesis parameters to the automatic update unit 17 in order to create intermediate information that can be used to create a synthesized speech signal.

  The speech synthesizer 11 creates a synthesized speech signal based on the set of input text information and synthesis parameters. Alternatively, the speech synthesizer 11 creates a synthesized speech signal based on the intermediate information stored in the storage unit 3. Furthermore, the speech synthesizer 11 creates intermediate information in response to an instruction from the automatic update unit 17. For this purpose, the speech synthesis unit 11 includes a language processing unit 12, a prosody generation unit 13, a segment selection unit 14, and a waveform generation unit 15.

The language processing unit 12 converts the input text information into phonetic information. The phonetic information is information representing the reading of the original text included in the text information. For example, the phonetic information is information in which the reading of the original text is represented by katakana characters, and the accent position and the break position are added.
The language processing unit 12 reads a language dictionary stored in the storage unit 3 in order to convert the input text information into phonetic information. In the language dictionary, for example, various words assumed to appear in text information, readings of the words, parts of speech, and utilization forms are registered. Then, the language processing unit 12 uses the language dictionary, for example, to perform morphological analysis on the original sentence included in the text information, and order and read each word appearing in the original sentence, position of the accent, and position of the break To decide. At that time, the language processing unit 12 sets, for example, a position where a punctuation mark is set in the original text as a break position. A unit of a sentence obtained by dividing an original sentence at a position where a punctuation mark is set is referred to as an exhalation paragraph in this specification.
As the morphological analysis, the language processing unit 12 can use, for example, a method using dynamic programming or a hidden Markov model. Then, the language processing unit 12 creates phonetic information according to the order of each word, reading, accent position, and break position.
The language processing unit 12 outputs the phonetic information to the prosody generation unit 13. The language processing unit 12 may output phonetic information to the division processing unit 18 of the automatic update unit 17 in order to generate intermediate information.

The prosody generation unit 13 generates a target prosody for generating a synthesized speech based on the synthesis parameter received from the coincidence determination unit 10 and the phonetic information received from the language processing unit 12. For this purpose, the prosody generation unit 13 reads a plurality of prosody models from the storage unit 3. This prosodic model represents a position in which the voice is raised and a position in which the voice is lowered in time order. Then, the prosody generation unit 13 selects a prosody model that most closely matches the position of the accent indicated by the phonetic information from among the plurality of prosody models. Then, according to the selected prosodic model and synthesis parameters, the prosody generation unit 13 creates a target prosody by setting a position where the voice becomes higher or lower, a voice inflection, a pitch, and the like with respect to the phonetic information. To do. The target prosody includes a phoneme length and a pitch frequency for each phoneme as a unit for determining a speech waveform. Note that the phoneme can be, for example, one vowel or one consonant. In this embodiment, a syllable in which one or more vowels and consonants are combined is also included in the phoneme.
The prosody generation unit 13 outputs the target prosody to the segment selection unit 14.
When the speech synthesis unit 11 generates intermediate information in response to an instruction from the automatic update unit 17, the prosody generation unit 13 acquires a set of text information and synthesis parameters from the automatic update unit 17. Then, the prosody generation unit 13 creates a target prosody based on the text information acquired from the automatic update unit 17 and the synthesis parameter, and outputs the target prosody to the segment selection unit 14.

The segment selection unit 14 and the waveform generation unit 15 create a synthesized speech signal by, for example, a phoneme connection method, a corpus base method, or a large-scale corpus base method.
The segment selection unit 14 selects, for each phoneme, a speech waveform closest to the phoneme length and pitch frequency of the target prosody from a plurality of speech waveforms registered in the speech waveform dictionary by pattern matching, for example. For this purpose, the segment selection unit 14 reads the speech waveform dictionary from the storage unit 13. The speech waveform dictionary records a plurality of speech waveforms and an identification number of each speech waveform. The voice waveform is, for example, a waveform signal extracted in units of phonemes from voice signals obtained by recording various voices in which one or more narrators read various texts.
Furthermore, in order to enable the segment selection unit 14 to connect the speech waveforms selected for each phoneme along the target prosody, the selected speech waveform and the waveform pattern of the corresponding phoneme indicated in the target prosody May be calculated as waveform conversion information.
The segment selection unit 14 creates waveform generation information including the identification number of the speech waveform selected for each phoneme. The waveform generation information may further include waveform conversion information. Then, the element selection unit 14 outputs the waveform generation information to the waveform generation unit 15. The element selection unit 14 outputs the waveform generation information to the automatic update unit 17 in order to store the waveform generation information as intermediate information.
Further, when the speech synthesis unit 11 creates intermediate information based on the combination of text information and synthesis parameters received from the automatic update unit 17, the segment selection unit 14 outputs the created waveform generation information to the automatic update unit 17. .

The waveform generation unit 15 generates a synthesized speech signal based on the waveform generation information. For this purpose, the waveform generation unit 15 reads the speech waveform signal corresponding to the speech waveform identification number of each phoneme included in the waveform generation information received from the segment selection unit 14 or the control unit 16 from the storage unit 3. And the waveform generation part 15 produces a synthetic | combination audio | voice signal by connecting each audio | voice waveform signal continuously. When the waveform conversion information is included in the waveform generation information, the waveform generation unit 15 corrects each speech waveform signal according to the waveform conversion information obtained for the corresponding phoneme and continuously connects the speech waveform signals. By doing so, a synthesized speech signal is created.
The waveform generation unit 15 outputs the synthesized speech signal to the output unit 5.

When the registered text information and registered composite parameter pair that matches the input text information and composite parameter pair is registered in the intermediate information table, the control unit 16 corresponds to the registered text information and registered parameter pair. The registered intermediate information is read from the storage unit 3. When the waveform generation information is included in the registration intermediate information, the control unit 16 outputs the waveform generation information included in the registration intermediate information to the waveform generation unit 15. When the registered intermediate information includes phonetic information and does not include waveform generation information, the control unit 16 outputs the phonetic information included in the registered intermediate information to the prosody generating unit 13.
Thereby, the speech synthesizer 11 can omit at least a part of the processing related to the generation of the phonetic information, the target prosody, and the waveform generation information. As a result, the speech synthesizer 11 can shorten the processing time required to create the synthesized speech signal.

  The automatic updating unit 17 causes the speech synthesis unit 11 to create intermediate information based on derived text information obtained by modifying the input text information or derived synthetic parameters obtained by changing at least one value of the input synthetic parameters. For this purpose, the automatic update unit 17 includes a division processing unit 18 and a parameter adjustment unit 19.

When the division processing unit 18 determines that the combination of registered text information and registered synthesis parameter that matches the set of text information and synthesis parameter input by the match determination unit 10 is not registered in the intermediate information table, the input Received text information is received from the match determination unit 10. Then, the division processing unit 18 divides the original text included in the input text information into predetermined units so that it can be easily used as intermediate information.
For example, the division processing unit 18 detects each punctuation mark included in the original sentence, and divides the original sentence so that a section from one punctuation mark to the next punctuation mark becomes a unit of one sentence. Alternatively, the division processing unit 18 detects a breakable position where a punctuation mark can be added in the original text regardless of the punctuation mark attached to the original text, and a section between adjacent breakable positions is a unit of one sentence. The original text may be divided so that In order to detect a breakable position, the division processing unit 18 may detect a phrase boundary by performing morphological analysis, for example, and may set the phrase boundary as a breakable position.

  The division processing unit 18 may receive the phonetic information created by the language processing unit 12 based on the input text information, and may divide the original text based on the phonetic information. In this case, for example, the division processing unit 18 divides the original text at a break position set in the phonetic information. Or the division | segmentation process part 18 may divide | segment an original sentence in the delimitation position which can delimit an original sentence other than the delimitation position set to the phonetic information. Also in this case, the division processing unit 18 detects a break position by, for example, performing a morphological analysis to detect a phrase boundary.

The division processing unit 18 creates text information including any one of sentence units created by dividing the original sentence as derived text information.
Further, the division processing unit 18 may create a plurality of types of sentence units for one original sentence by dividing the original sentence at different positions. For example, it is assumed that the original text included in the text information is “Notice of traffic congestion to the driver traveling in the direction of Nagoya.” In this case, the division processing unit 18 may divide the original into two units of “to a driver traveling in the direction of Nagoya” and “a notification of traffic jam”. Alternatively, the division processing unit 18 may divide the original into three units: “To Nagoya,” a unit, “To the driving driver,” and “Congestion notification.” Furthermore, the division processing unit 18 may divide the original text into two units, such as “to the Nagoya area” and “is a traffic jam notification to the driving driver”. In this case, the division processing unit 18 may create derivative text information for each sentence unit.

The parameter adjustment unit 19 creates a derived synthesis parameter by changing at least one of the input synthesis parameters. For example, it is assumed that the speech speed and pitch included in the synthesis parameter are represented by “1” to “5”, respectively. In this case, if the speech speed and pitch of the input synthesis parameter are “3”, the parameter adjustment unit 19 corrects the speech speed to “1”, “2”, “4”, or “5”. Alternatively, the derived synthesis parameter is created by modifying the pitch to '1', '2', '4' or '5'.
Then, the automatic update unit 17 passes the combination of the derived text information and the derived synthesis parameter to the speech synthesis unit 11, and causes the speech synthesis unit 11 to create phonetic information and waveform generation information for the combination of the derived text information and the derived synthesis parameter. .
It should be noted that a combination of registered text information and a registered composite parameter that matches a set of derived text information and a derived composite parameter may already be registered in the intermediate information table. In such a case, the automatic updating unit 17 may not cause the speech synthesis unit 11 to create phonetic information and waveform generation information for the combination of the derived text information and the derived synthesis parameter.

When the automatic update unit 17 receives the phonetic information and the waveform generation information for the set of the derived text information and the derived synthesis parameter from the speech synthesis unit 11, the automatic update unit 17 sets the set of the derived text information and the derived synthesis parameter together with the phonetic information and the waveform generation information. Add to the intermediate information table. As described above, the automatic update unit 17 updates the intermediate information table stored in the storage unit 3, so that intermediate information that can be used to create a synthesized speech signal for a set of text information and synthesis parameters input thereafter is stored. Increase the number. That is, the combination of the derived text information and the derived synthesis parameter becomes a set of new registered text information and the registered synthesis parameter. The phonetic information and the waveform generation information created for the combination of the derived text information and the derived synthesis parameter become registration intermediate information corresponding to the new registered text information and registered synthesis parameter group.
Note that the automatic update unit 17 may create intermediate information by changing only one of the input text information and the synthesis parameter. That is, either the derived text information or the derived synthesis parameter may be the same as the input text information or the input synthesis parameter. Further, the automatic updating unit 17 may register only one of the phonetic information and the waveform generation information created for the combination of the derived text information and the derived synthesis parameter as registered intermediate information in the intermediate information table.

FIG. 3 is a diagram illustrating an example of the intermediate information table. Each row of the intermediate information table 300 corresponds to one registered intermediate information. In order from the left, each column stores an identification number of each registered intermediate information, a synthesis parameter, text information, phonetic information, and waveform generation information. In this example, intermediate information corresponding to the text information and synthesis parameters input in line 301 is registered, while lines 302 and 303 are created based on the input text information and derived synthesis parameters. Intermediate information is registered.
In this example, speech speed and pitch are defined as synthesis parameters. Therefore, at least one of the speech speed or pitch of the derived synthesis parameter is different from the original speech speed or pitch shown in the row 301. Note that the synthesis parameters may include other parameters such as intonation, volume, and voice pitch.

In addition, for example, if the synthesized speech is Japanese, the text information is represented by a combination of kana characters, kanji and punctuation marks as in a general sentence. For example, in this example, the text information includes a sentence “I want to say to the customer”.
The phonetic information is expressed in accordance with, for example, the Japan Electronics and Information Technology Industries Association Standard TT-6004 (Automotive Speech Synthesis Symbol Standard). Moreover, the phonetic information may be described according to other notation formats in which the accent position and the break position in the sentence included in the text information and the reading of each phoneme can be understood. In this example, the sentence and phonetic information included in the derived text information are the same as the sentence and phonetic information included in the input text information.
In the column in which the waveform generation information is registered, the identification number of the speech waveform used for each phoneme and the waveform conversion information depending on the case are described along the time series. It should be noted that an appropriate one may be selected as the description format of the identification number and the waveform conversion information.

FIG. 4 is a diagram illustrating another example of the intermediate information table. Each row of the intermediate information table 400 corresponds to one registered intermediate information. In order from the left, each column stores an identification number of each registered intermediate information, a synthesis parameter, text information, phonetic information, and waveform generation information. In this example, the input text information includes the original text “This is a traffic jam notification to the driver traveling in the direction of Nagoya.” In lines 401 and 402, text information divided by punctuation marks added to the sentence in the input text information and intermediate information corresponding to the input synthesis parameters are registered. On the other hand, the lines 403 and 404 are created based on the derived text information created by dividing the original text at a position different from the position of the punctuation mark added in the inputted text information and the inputted synthesis parameter. Intermediate information is registered.
In this example, in the line 403, the sentence “To Nagoya,” is registered as the derived text information, while in the line 404, the message “Notice of traffic congestion to the driving driver”. Is registered.
In this way, by registering the intermediate information for the derived text information generated by variously changing the unit for dividing one sentence into the intermediate information table, the text information input thereafter and the intermediate information table can be stored in the intermediate information table. Probability of matching registered text information increases. Therefore, the speech synthesizer 1 can increase the possibility that the time required for creating the synthesized speech can be shortened.

FIG. 5 shows an operation flowchart of the speech synthesis process executed by the processing unit 4 of the speech synthesizer 1.
The processing unit 4 acquires text information and synthesis parameters via the input unit 2 (step S101).
The coincidence determination unit 10 is a combination of registered text information or phonetic information and a registered synthesis parameter in which an input text information or a set of phonetic information represented by the text information and a synthesis parameter is registered in an intermediate information table. It is determined whether or not it matches any one of these (step S102).
When the set of the input text information or phonetic information and the synthesis parameter matches any of the set of registered text information or phonetic information and the registered synthesis parameter (step S102—Yes), the match determination unit 10 An identification number of a set of the registered text information or phonetic information that matches and the registered synthesis parameter is passed to the control unit 16. The control unit 16 reads the registered intermediate information corresponding to the identification number received from the match determination unit 10 from the storage unit 3 with reference to the intermediate information table (step S103). Then, the control unit 16 passes the registered intermediate information to the speech synthesis unit 11.
The voice synthesizer 11 creates a synthesized voice signal using the registered intermediate information (step S104). Then, the voice synthesizer 11 outputs the synthesized voice signal to the speaker 6 via the output unit 5.

On the other hand, when the set of the input text information or the phonetic information and the synthesis parameter does not match any of the set of the registered text information or the phonetic information and the registered synthesis parameter (step S102—Yes), the match determination unit 10 is input. The combination of the text information and the synthesis parameter is passed to the speech synthesis unit 11. Then, the speech synthesizer 11 creates a synthesized speech signal based on the input text information and synthesis parameters (step S105). The voice synthesizer 11 outputs the synthesized voice signal to the speaker 6 via the output unit 5. The speech synthesizer 11 passes the phonetic information and waveform generation information created in the middle to create a synthesized speech signal as intermediate information to the automatic update unit 17 together with the input text information and synthesis parameters.
The automatic update unit 17 registers the input text information, the synthesis parameter, and the created phonetic information and waveform generation information in the intermediate information table as registered text information, registered synthesis parameter, and registered intermediate information, respectively (step S106). .

Further, the automatic update unit 17 receives the input text information and the synthesis parameter from the coincidence determination unit 10. Then, the division processing unit 18 of the automatic update unit 17 divides the original text included in the text information into a plurality of predetermined sentence units, and creates derivative text information for each unit of the sentences (step S107).
The parameter adjustment unit 19 of the automatic updating unit 17 creates a derived synthesis parameter by correcting at least one parameter included in the input synthesis parameter (step S108).

The automatic update unit 17 outputs the derived text information and the derived synthesis parameter to the speech synthesis unit 11. Note that the timing at which the derived text information and the derived synthesis parameter are passed from the automatic update unit 17 to the speech synthesis unit 11 is that the speech synthesis unit 11 does not create a synthesized speech signal in order to reduce the load on the speech synthesis unit 11. It is preferred that Therefore, for example, the automatic update unit 17 receives a notification from the waveform generation unit 15 of the speech synthesis unit 11 that the creation of the synthesized speech signal for the combination of the input text information and the input synthesis parameter is completed, and then derives The text information and the derived synthesis parameters are passed to the speech synthesis unit 11. Then, the speech synthesizer 11 creates phonetic information and waveform generation information based on the derived text information and the derived synthesis parameter (step S109).
The speech synthesis unit 11 passes the created phonetic information and waveform generation information to the automatic update unit 17 as intermediate information. Then, the automatic update unit 17 registers the derived text information, the derived synthesis parameter, and the created phonetic information and waveform generation information in the intermediate information table as registered text information, registered synthesis parameter, and registered intermediate information, respectively (Step). S110).
After step S104 or step S110, the processing unit 4 ends the speech synthesis process.
Note that a part of the original text included in the input text information and a set of registered text information and a set of registered composite parameters that match a set of input composite parameters may be registered in the intermediate information table. In such a case, the processing unit 4 executes the processes of step S103 and step S104 for the registered text information and the registered composite parameter set that match a part of the original text and the input composite parameter set. On the other hand, the processing unit 4 executes the processes of steps S105 to S110 for the other parts of the original text.
Further, the processing unit 4 may exchange the order of the process of step S107 and the process of step S108, or may omit either one of the process of step S107 or the process of step S108.

  As described above, this speech synthesizer is configured to synthesize speech based on the input text information and synthesis parameters unless intermediate information corresponding to the set of input text information and synthesis parameters is stored. Create a signal. The speech synthesizer stores waveform generation information and the like created during the creation of the synthesized speech signal together with a set of text information and synthesis parameters input as intermediate information. Further, the speech synthesizer creates intermediate information such as waveform generation information for the derived text information obtained by correcting the input text information and the derived synthesis parameter obtained by correcting a part of the synthesis parameter. The speech synthesizer stores the intermediate information together with the derived text information and the derived synthesis parameter so that it can be used for speech synthesis. For this reason, this speech synthesizer is required to create a synthesized speech signal under a synthesis condition that is different from the synthesized condition of the synthesized speech signal for an original text that has at least a part of the original text of the synthesized speech signal once created. In this case, the time required for speech synthesis can be shortened.

Next, a speech synthesizer according to the second embodiment will be described.
The speech synthesizer according to the second embodiment checks the number of uses of each piece of intermediate information registered in the intermediate information table. The speech synthesizer then deletes the intermediate information that is used less frequently and the corresponding registered text information and registered synthesis parameter set. As a result, this speech synthesizer shortens the time required for searching for a set of registered text information and registered synthesis parameters that match the set of input text information and synthesis parameters.

FIG. 6 is a schematic configuration diagram of the processing unit 41 of the speech synthesizer according to the second embodiment. The processing unit 41 includes a coincidence determination unit 10, a voice synthesis unit 11, a control unit 16, and an automatic update unit 17. The automatic update unit 17 includes a division processing unit 18, a parameter adjustment unit 19, and a survival determination unit 20.
In FIG. 6, each component of the processing unit 41 is assigned the same reference number as the corresponding component of the processing unit 4 of the speech synthesizer 1 according to the first embodiment shown in FIG. 2. The processing unit 41 according to the second embodiment is different from the processing unit 4 according to the first embodiment in that the automatic update unit 17 includes a survival determination unit 20.
Therefore, hereinafter, the difference between the processing unit 41 and the processing unit 4 according to the first embodiment will be described. For other components of the speech synthesizer according to the second embodiment, refer to FIGS. 1 and 2 and the description of the relevant part of the first embodiment.

  In this embodiment, the intermediate information table includes the latest date and time when the number of times used to create a synthesized speech signal for each set of registered text information and registered synthesis parameters, and the synthesized speech signal after the time of the check. Further record the number of times used for creation.

  FIG. 7 is a diagram illustrating an example of an intermediate information table according to the second embodiment. Each row of the intermediate information table 700 corresponds to one registered intermediate information. In order from the left, each column stores an identification number of each registered intermediate information, synthesis parameters, text information, phonetic information, waveform generation information, number of uses, and check date and time. For example, for the intermediate information registered in the row 701, it is shown that the usage count is checked at 0:00 on August 25, 2010. The intermediate information registered in the row 701 indicates that it has been used 132 times since the previous check. Similarly, it is shown that the number of uses of the intermediate information registered in the row 702 has been checked at 0:00 on August 20, 2010. The intermediate information registered in the row 702 indicates that it has been used 16 times since the previous check.

In this embodiment, when any registered intermediate information is used to create a synthesized speech signal, the control unit 16 increments the number of uses for the registered intermediate information stored in the intermediate information table. That is, the number of times of using the registered intermediate information corresponding to the set of registered text information and registered composite parameter determined to match the set of text information and composite parameter input by the match determination unit 10 is increased by one.
Further, when the intermediate update information is newly registered in the intermediate information table, the automatic updating unit 17 sets the number of uses for the intermediate information to “0” and the check date as the registration date.

The existence determination unit 20 of the automatic update unit 17 erases the registered intermediate information that is not used so much in the most recent predetermined period among the registered intermediate information registered in the intermediate information table.
FIG. 8 is an operation flowchart of the registration intermediate information existence determination process executed by the existence determination unit 20.
The survival determination unit 20 refers to the intermediate information table stored in the storage unit 3 regularly (for example, every day at midnight), and identifies registered intermediate information that has passed a predetermined period from the previous check date (step) S201). Alternatively, the survival determination unit 20 may specify registration intermediate information that has passed a predetermined period from the previous check date and time at a specific timing (for example, when the speech synthesizer is activated or terminated). This predetermined period is determined according to the frequency with which the speech synthesizer itself is used, the storage capacity of the storage unit 3 or the processing speed of the processor included in the processing unit 41. For example, the higher the frequency of use of the speech synthesizer, the shorter the predetermined period is set, while the larger the storage capacity of the storage unit 3 or the faster the processing speed of the processor included in the processing unit 41, the longer the predetermined period is set. Is done. For example, the predetermined period is set to one week, one month, six months, or one year.

The survival determination unit 20 determines whether or not the number of times of use of the specified registered intermediate information is less than a predetermined threshold (step S202). The predetermined threshold is set, for example, from about once to several times. If the number of times of use is less than a predetermined threshold (step S202—Yes), the intermediate information corresponding to the number of times of use is hardly used for creating a synthesized speech signal. Therefore, the survival determination unit 20 deletes from the intermediate information table the registered intermediate information whose usage count is less than the predetermined threshold and the corresponding registered text information and registered composite parameter set (step S203).
On the other hand, when the number of times of use is equal to or greater than a predetermined threshold (step S202-No), the registered intermediate information is used to create a synthesized speech signal. Therefore, it is preferable that the speech synthesizer keeps such intermediate information. Accordingly, the survival determination unit 20 corrects the check date and time recorded in the intermediate information table for the registered intermediate information whose number of uses is equal to or greater than a predetermined threshold to the date and time when this check was performed. Further, the survival determination unit 20 resets the number of uses for the checked intermediate information to “0” (step S204).
When there are a plurality of pieces of registered intermediate information that have passed a predetermined period from the previous check date and time, the survival determination unit 20 executes the processes of steps S202 to S204 for each of the registered intermediate information. And the survival determination part 20 complete | finishes a survival determination process, if the process of step S202-S204 is complete | finished about all the registration intermediate information.

  Thereby, this speech synthesizer can erase the intermediate information with low usage frequency from the storage unit. Therefore, the speech synthesizer can suppress an unnecessary increase in the number of intermediate information registered in the intermediate information table. Therefore, this speech synthesizer can reduce the time required for searching for registered intermediate information corresponding to the set of registered text information and registered synthesis parameter that matches the set of input text information and synthesis parameter. Furthermore, since this speech synthesizer determines whether or not to delete each registered intermediate information based on the number of times of use in the latest fixed period, it is possible to appropriately select the registered intermediate information to be deleted.

Next, a speech synthesizer according to a third embodiment will be described.
The speech synthesizer according to the third embodiment has means for editing registered intermediate information. As a result, when the intermediate information registered in the intermediate information table is inappropriate, the speech synthesizer can manually delete the registered intermediate information or correct the registered intermediate information.

FIG. 9 is a schematic configuration diagram of the processing unit 42 of the speech synthesizer according to the third embodiment. The processing unit 42 includes a match determination unit 10, a voice synthesis unit 11, a control unit 16, an automatic update unit 17, and an editing unit 21.
In FIG. 9, each component of the processing unit 42 is assigned the same reference number as that of the corresponding component of the processing unit 4 of the speech synthesizer 1 according to the first embodiment shown in FIG. The processing unit 42 according to the third embodiment is different from the processing unit 4 according to the first embodiment in that the editing unit 21 is included.
Therefore, hereinafter, the difference between the processing unit 42 and the processing unit 4 according to the first embodiment will be described. For other components of the speech synthesizer according to the third embodiment, refer to FIGS. 1 and 2 and the description of the relevant part of the first embodiment.

In this embodiment, the input unit 2 includes a pointing device such as a keyboard or a mouse and a display, for example. Alternatively, the input unit 2 may have a touch panel display.
When the processing unit 42 receives an operation signal indicating that the registered intermediate information is to be edited from the input unit 2, the processing unit 42 activates the editing unit 21.

The editing unit 21 displays, for example, a menu for the user to select registration intermediate information to be edited or an operation button on the display of the input unit 2. The editing unit 21 displays on the display an operation button for deleting the registered intermediate information to be edited or a text box for correcting the phonetic information included in the registered intermediate information.
When the editing unit 21 acquires the identification number of the registration intermediate information to be edited from the keyboard or the like of the input unit 2, the editing unit 21 registers the registration intermediate information corresponding to the identification number, the corresponding registered text information, and the registration composition parameter from the storage unit 3. Is read. Then, the editing unit 21 displays the registered text information and the registered synthesis parameter and the phonetic information or waveform generation information included in the registered intermediate information on the display of the input unit 2.
When the editing unit 21 receives an operation signal for deleting the registered intermediate information to be edited and the identification number of the registered intermediate information from the keyboard or the like of the input unit 2, the editing unit 21 deletes the registered intermediate information from the intermediate information table.
Alternatively, when the editing unit 21 receives an operation signal for correcting a part of the registered intermediate information to be edited, for example, a part of the phonetic information and the identification number of the registered intermediate information from the keyboard or the like of the input unit 2, According to the operation signal, the phonetic information of the registered intermediate information specified by the identification number is corrected. The editing unit 21 inputs the corrected phonetic information to the speech synthesis unit 11 to obtain waveform generation information corresponding to the corrected phonetic information.

  Thus, since this speech synthesizer has means for editing already registered intermediate information, for example, it is based on phonetic information represented by a reading different from the original reading of the sentence indicated in the text information. Like intermediate information, inappropriate intermediate information can be edited. Therefore, this speech synthesizer can leave only appropriate intermediate information.

According to the modification of each embodiment described above, the parameter adjustment unit of the automatic update unit replaces the proper noun included in the input text information or the derived text information with another proper noun of the same type as the proper noun. Thus, another derived text information may be created. For example, if the text information contains the sentence “To the driver driving in the direction of Nagoya,” the parameter adjustment unit will change the proper name of the place name to “Nagoya”. The noun “Tokyo” or “Kyoto” may be substituted. In order to perform such proper noun replacement, for example, the word dictionary stored in the storage unit also includes identification information indicating the type of proper noun. The division processing unit notifies the parameter adjustment unit of the part of speech of each word included in the text information obtained as a result of the morphological analysis.
As a result, the automatic update unit can cause the speech synthesis unit to create intermediate information for more derived text information, so intermediate information can be used to create a synthesized speech signal for the next set of input text information and synthesis parameters. The possibility that it can be used can be further increased.

  According to another modification, the automatic update unit may cause the speech synthesis unit to create a synthesized speech signal corresponding to the derived text information and the derived synthesis parameter. In this case, the automatic update unit may receive the synthesized speech signal from the speech synthesis unit and store the synthesized speech signal in the storage unit as registered intermediate information. Similarly, the automatic updating unit may store the synthesized speech signal created based on the input text information and the input synthesis parameter in the storage unit as registered intermediate information. As described above, when the storage unit stores the synthesized speech signal, the registered text information and the set of registered synthesis parameters that match the set of text information and the synthesis parameters input thereafter are registered in the intermediate information table. The speech synthesizer can also omit the processing of the waveform generator. Therefore, the speech signal device can further reduce the time required for speech synthesis.

Furthermore, in each of the embodiments described above, the speech synthesizer may acquire the importance of the input text information together with the input text information and the combination parameter through the input unit. For example, the importance is set to two levels or three levels or more. For example, the greater the numerical value representing the importance, the higher the importance of the input text information.
The automatic updating unit may cause the speech synthesis unit to create intermediate information for the combination of the derived text information and the derived synthesis parameter only when the importance is equal to or greater than a predetermined importance level threshold. For example, when the importance level is set to either “0” or “1”, the importance level threshold is set to “1”. In addition, when the importance is set to any one of three or more levels such as “0” to “n” (where n is an integer of 2 or more), the importance threshold is set to any one of “1” to “n”. Is set.

  Furthermore, when the importance is set to any one of the three levels, the automatic update unit may determine intermediate information to be registered in the intermediate information table according to the importance. For example, if the level of importance is the lowest (for example, “0”), the automatic update unit does not register the intermediate information created for the set of input text information and composite parameters in the intermediate information table. If the importance level is in an intermediate stage (for example, “1”), the automatic updating unit registers the intermediate information created for the set of input text information and synthesis parameter in the intermediate information table. However, the automatic updating unit does not register the intermediate information about the set of the derived text information and the derived synthesis parameter in the intermediate information table. If the level of importance is the highest (for example, “2”), the automatic update unit registers the intermediate information created for the set of input text information and composite parameters in the intermediate information table. Further, the automatic update unit also registers intermediate information about the combination of the derived text information and the derived synthesis parameter in the intermediate information table. Furthermore, when the importance is set to any of four or more levels, the automatic update unit sets a larger number of combinations of the derived text information and the derived synthesis parameter as the importance is “2” or higher and the value is higher. You may register the intermediate information about.

  In this modification, the automatic updating unit records the importance level together with each piece of intermediate information in the intermediate information table. For example, for the intermediate information created for the combination of the input text information and the synthesis parameter, the automatic update unit records the importance input together with the combination in the intermediate information table. In addition, the automatic update unit is also set for the original input text information and the composite parameter set for the intermediate information obtained for the derived text information and the derived composite parameter set created from the input text information and the composite parameter set. Record the importance. Alternatively, the automatic update unit sets, in the intermediate information table, an importance level lower than the importance level set for the original input text information and the synthesis parameter set for the intermediate information obtained for the set of the derived text information and the derived synthesis parameter. It may be recorded.

  Furthermore, when the automatic update unit has a survival determination unit, the survival determination unit may determine whether to delete the registered intermediate information according to the importance recorded in the intermediate information table. For example, the survival determination unit does not delete the registration intermediate information for the registration intermediate information whose importance is higher than the second importance threshold even if the number of uses is less than a predetermined threshold. The second importance level threshold is set to a value higher than the above importance level threshold, for example. Alternatively, the survival determination unit is used for comparison with the number of times of use of the registered intermediate information when the importance recorded for the registered intermediate information is the lowest of the levels of importance in which the intermediate information is registered. The threshold value may be set higher than a normally used threshold value.

  FIG. 10 is a diagram showing another example of the intermediate information table according to this modification. Each row of the intermediate information table 1000 corresponds to one registered intermediate information. In order from the left, each column stores an identification number, synthesis parameter, text information, phonetic information, waveform generation information, importance, number of uses, and check date and time of each registered intermediate information. For example, the intermediate information registered in the row 1001 indicates that the importance is “5”. In addition, it is indicated that this registration intermediate information has not been used once since the previous check. Similarly, the importance level of the intermediate information registered in the row 1002 is “1”. This registration intermediate information is also shown to have never been used since the previous check. In this case, if the second importance threshold is “2”, for example, and the threshold used for comparison with the number of uses is “1” or more, the intermediate information registered in the row 1002 has both the importance and the number of uses. Since it is less than the threshold value, it is deleted by the survival determination unit. On the other hand, the intermediate information registered in the row 1001 is not deleted because the importance is higher than the second importance threshold.

  Furthermore, a computer program that causes a computer to realize each function of the processing unit of the speech synthesizer according to each of the above embodiments is recorded on a computer-readable medium, for example, a magnetic recording medium, an optical recording medium, or a semiconductor memory. It may be provided in the form.

  All examples and specific terms listed herein are intended for instructional purposes to help the reader understand the concepts contributed by the inventor to the present invention and the promotion of the technology. It should be construed that it is not limited to the construction of any example herein, such specific examples and conditions, with respect to showing the superiority and inferiority of the present invention. Although embodiments of the present invention have been described in detail, it should be understood that various changes, substitutions and modifications can be made thereto without departing from the spirit and scope of the present invention.

The following supplementary notes are further disclosed regarding the embodiment described above and its modifications.
(Appendix 1)
An input unit that obtains input text information including an original sentence that is a source of the synthesized speech signal and an input synthesis condition for creating the synthesized speech signal;
A set of registered text information and registered synthesis condition, and a set of registered intermediate information including registered phonetic information, which is generated in the middle of creating a synthesized speech signal corresponding to the set of registered text information and the registered synthesis condition A storage unit for storing at least one of
Whether the set of the input text information or the input phonetic information represented by the input text information and the input synthesis condition matches the registered text information or the set of the registered phonetic information and the registered synthesis condition A match determination unit for determining whether or not,
When the set of the input text information or the input phonetic information and the input synthesis condition matches either the registered text information or the set of the registered phonetic information and the registered synthesis condition, the registered text information or the A synthesized speech signal is created using the registered intermediate information corresponding to a set of registered phonetic information and the registered synthesis condition, while a set of the input text information or the input phonetic information and the input synthesis condition is A speech synthesizer that creates a synthesized speech signal based on the input text information and the input synthesis condition if it does not match any of the set of the registered text information or the registered phonetic information and the registered synthesis condition;
When the set of the input text information or the input phonetic information and the input synthesis condition does not match any of the registered text information or the set of the registered phonetic information and the registered synthesis condition, the input text information or the A set of derived text information and a derived synthesis condition is created by modifying the input synthesis condition, and the speech synthesis unit based on the derived text information and the derived synthesis condition set, the derived text information and the derived synthesis condition Is a set of derived intermediate information created by executing up to the middle stage of generating a synthesized speech signal as one of a set of the registered text information, the set of registered synthesis conditions, and the registered intermediate information. An automatic update unit to be stored in the storage unit;
A speech synthesizer.
(Appendix 2)
The automatic update unit creates a plurality of second original sentences by dividing the original sentence included in the input text information into predetermined units, and includes any one of the plurality of second original sentences. The speech synthesizer according to appendix 1, further comprising a division processing unit that creates derived text information.
(Appendix 3)
The speech synthesis apparatus according to appendix 2, wherein the predetermined unit is an exhalation paragraph set in the original text.
(Appendix 4)
The speech synthesizer according to appendix 2, wherein the predetermined unit is a unit of a sentence delimited by at least one position where a punctuation mark can be set in the original sentence.
(Appendix 5)
The speech synthesizer includes a language processing unit that obtains a first position that delimits the original text by performing morphological analysis on the original text,
The division processing unit creates one original sentence of the plurality of second original sentences by dividing the original sentence at the first position, and can set punctuation marks different from the first position. The speech synthesizer according to appendix 2, wherein another original text is created from the plurality of second original texts by dividing the original text at a second position.
(Appendix 6)
The speech synthesizer according to appendix 1, wherein the automatic update unit creates the derived text information by replacing a proper noun in the original sentence with another proper noun of the same type as the proper noun.
(Appendix 7)
The automatic update unit deletes the registration intermediate information that is used to create a synthesized speech signal within the most recent predetermined period from the registration intermediate information, the number of times being less than a predetermined threshold. The speech synthesizer according to claim 1.
(Appendix 8)
The input unit acquires the importance of the input text information together with the input text information,
The automatic update unit determines a second importance of the intermediate information according to the importance of the input text information, and determines the second importance as a set of the derived text information and the derived composition condition. And associated with the set of derived intermediate information and stored in the storage unit,
And among the registration intermediate information, the number of times used for creating a synthesized speech signal within the most recent predetermined period is less than a predetermined threshold, and the second importance is less than the first importance threshold. The speech synthesizer according to any one of appendices 1 to 6, wherein certain registration intermediate information is deleted.
(Appendix 9)
The input unit acquires the importance of the input text information together with the input text information, and the automatic update unit is configured to obtain the derived text information when the importance of the input text information is greater than or equal to a second importance threshold. And the combination of the derived synthesis condition and the derived intermediate information are stored in the storage unit, and when the importance of the input text information is less than the second importance threshold, the derived text information and The speech synthesizer according to any one of appendices 1 to 6, wherein the set of the derived synthesis condition and the set of the derived intermediate information is not stored in the storage unit.
(Appendix 10)
Obtaining input text information including the original text that is the source of the synthesized speech signal and input synthesis conditions for creating the synthesized speech signal;
The set of the input text information or the input phonetic information represented by the input text information and the input synthesis condition is at least one registered text information or registered phonetic information and the registration synthesis condition stored in the storage unit. Whether it matches any of the set of
When the set of the input text information or the input phonetic information and the input synthesis condition matches either of the registered text information or the set of the registered phonetic information and the registered synthesis condition, stored in the storage unit, In addition, a synthesized speech signal is created using registration intermediate information generated in the middle of creating a synthesized speech signal corresponding to the set of the registered text information or the registered phonetic information and the registered synthesis condition, When the set of the input text information or the input phonetic information and the input synthesis condition does not match any of the registered text information or the set of the registered phonetic information and the registered synthesis condition, the input text information and Create a synthesized speech signal based on the input synthesis conditions,
When the set of the input text information or the input phonetic information and the input synthesis condition does not match any of the registered text information or the set of the registered phonetic information and the registered synthesis condition, the input text information or the A set of derived text information and a derived synthesis condition is created by modifying the input synthesis condition, and a synthesized speech signal is generated based on the derived text information and the derived synthesis condition set, the derived text information, and the derived synthesis condition. A set of derived intermediate information created in the course of creating is stored in the storage unit as one of a set of the registered text information, the set of registered synthesis conditions, and the registered intermediate information.
A speech synthesis method.
(Appendix 11)
A computer program for speech signal synthesis that causes a computer to create a synthesized speech signal,
A set of input text information including an original sentence that is a source of a synthesized speech signal or input phonogram information represented by the input text information and an input synthesis condition for creating the synthesized speech signal is stored in a storage unit. Determining whether or not it matches any one of the set of at least one registered text information or registered phonetic information and registered synthesis condition;
When the set of the input text information or the input phonetic information and the input synthesis condition matches any of the registered text information or the set of the registered phonetic information and the registered synthesis condition, it is stored in the storage unit. And creating a synthesized speech signal using registered intermediate information generated in the middle of creating a synthesized speech signal corresponding to a set of the registered text information or the registered phonetic information and the registered synthesis condition, On the other hand, when the set of the input text information or the input phonetic information and the input synthesis condition does not match any of the registered text information or the set of the registered phonetic information and the registered synthesis condition, the input text information And a synthesized speech signal based on the input synthesis condition,
When the set of the input text information or the input phonetic information and the input synthesis condition does not match any of the registered text information or the set of the registered phonetic information and the registered synthesis condition, the input text information or the A set of derived text information and a derived synthesis condition is created by modifying the input synthesis condition, and a synthesized speech signal is generated based on the derived text information and the derived synthesis condition set, the derived text information, and the derived synthesis condition. A set of derived intermediate information created in the course of creating is stored in the storage unit as one of a set of the registered text information, the set of registered synthesis conditions, and the registered intermediate information.
A computer program that causes a computer to execute.

DESCRIPTION OF SYMBOLS 1 Speech synthesizer 2 Input part 3 Storage part 4, 41, 42 Processing part 5 Output part 6 Speaker 10 Match determination part 11 Speech synthesizer 12 Language processing part 13 Prosody generation part 14 Segment selection part 15 Waveform generation part 16 Control part 17 Automatic Update Unit 18 Division Processing Unit 19 Parameter Adjustment Unit 20 Survival Determination Unit 21 Editing Unit

Claims (7)

An input unit that obtains input text information including an original sentence that is a source of the synthesized speech signal and an input synthesis condition for creating the synthesized speech signal;
A set of registered text information and registered synthesis condition, and a set of registered intermediate information including registered phonetic information, which is generated in the middle of creating a synthesized speech signal corresponding to the set of registered text information and the registered synthesis condition A storage unit for storing at least one of
Whether the set of the input text information or the input phonetic information represented by the input text information and the input synthesis condition matches the registered text information or the set of the registered phonetic information and the registered synthesis condition A match determination unit for determining whether or not,
When the set of the input text information or the input phonetic information and the input synthesis condition matches either the registered text information or the set of the registered phonetic information and the registered synthesis condition, the registered text information or the A synthesized speech signal is created using the registered intermediate information corresponding to a set of registered phonetic information and the registered synthesis condition, while a set of the input text information or the input phonetic information and the input synthesis condition is A speech synthesizer that creates a synthesized speech signal based on the input text information and the input synthesis condition if it does not match any of the set of the registered text information or the registered phonetic information and the registered synthesis condition;
When the set of the input text information or the input phonetic information and the input synthesis condition does not match any of the registered text information or the set of the registered phonetic information and the registered synthesis condition, the input text information or the A set of derived text information and a derived synthesis condition is created by modifying the input synthesis condition, and the speech synthesis unit based on the derived text information and the derived synthesis condition set, the derived text information and the derived synthesis condition Is a set of derived intermediate information created by executing up to the middle stage of generating a synthesized speech signal as one of a set of the registered text information, the set of registered synthesis conditions, and the registered intermediate information. An automatic update unit to be stored in the storage unit;
A speech synthesizer.   The automatic update unit creates a plurality of second original sentences by dividing the original sentence included in the input text information into predetermined units, and includes any one of the plurality of second original sentences. The speech synthesizer according to claim 1, further comprising a division processing unit that creates derived text information.   The speech synthesizer according to claim 2, wherein the predetermined unit is an exhalation paragraph set in the original text. The speech synthesizer includes a language processing unit that obtains a first position that delimits the original text by performing morphological analysis on the original text,
The division processing unit creates one original sentence of the plurality of second original sentences by dividing the original sentence at the first position, and can set punctuation marks different from the first position. The speech synthesizer according to claim 2, wherein another original sentence is created among the plurality of second original sentences by dividing the original sentence at a second position.   5. The automatic update unit according to claim 1, wherein the automatic update unit deletes the registration intermediate information that is used to create a synthesized speech signal within the most recent predetermined period from the registration intermediate information that is less than a predetermined threshold. The speech synthesis apparatus according to any one of the above. The input unit acquires the importance of the input text information together with the input text information,
The automatic update unit determines a second importance of the intermediate information according to the importance of the input text information, and determines the second importance as a set of the derived text information and the derived composition condition. And associated with the set of derived intermediate information and stored in the storage unit,
And among the registration intermediate information, the number of times used for creating a synthesized speech signal within the most recent predetermined period is less than a predetermined threshold, and the second importance is less than the first importance threshold. The speech synthesizer according to any one of claims 1 to 4, wherein certain registered intermediate information is deleted. Obtaining input text information including the original text that is the source of the synthesized speech signal and input synthesis conditions for creating the synthesized speech signal;
The set of the input text information or the input phonetic information represented by the input text information and the input synthesis condition is at least one registered text information or registered phonetic information and the registration synthesis condition stored in the storage unit. Whether it matches any of the set of
When the set of the input text information or the input phonetic information and the input synthesis condition matches either of the registered text information or the set of the registered phonetic information and the registered synthesis condition, stored in the storage unit, In addition, a synthesized speech signal is created using registration intermediate information generated in the middle of creating a synthesized speech signal corresponding to the set of the registered text information or the registered phonetic information and the registered synthesis condition, When the set of the input text information or the input phonetic information and the input synthesis condition does not match any of the registered text information or the set of the registered phonetic information and the registered synthesis condition, the input text information and Create a synthesized speech signal based on the input synthesis conditions,
When the set of the input text information or the input phonetic information and the input synthesis condition does not match any of the registered text information or the set of the registered phonetic information and the registered synthesis condition, the input text information or the A set of derived text information and a derived synthesis condition is created by modifying the input synthesis condition, and a synthesized speech signal is generated based on the derived text information and the derived synthesis condition set, the derived text information, and the derived synthesis condition. A set of derived intermediate information created in the course of creating is stored in the storage unit as one of a set of the registered text information, the set of registered synthesis conditions, and the registered intermediate information.
A speech synthesis method.

Images