JP2007249022A - Speech synthesizer and speech synthesizing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech synthesizer capable of processing a text wherein a rule applied to a variable portion is described. <P>SOLUTION: The speech synthesizer includes an input means for inputting a rule set for speech synthesis and a character string to be substituted in a variable portion of the text, an extraction processing means of extracting a rule identifier from the text, a means of extracting information related to the rule from the character string input from the input means, and a variable portion processing means of selecting the rule to be applied according to the rule set and rule identifier and processing at least the variable portion based upon information related to the selected rule according to the rule. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a speech synthesizer and a speech synthesis method for processing text including control rules.

Devices such as car navigation systems use speech synthesis for response messages to users. When generating a response message using speech synthesis, it is common to prepare text including control tags such as reading and accent in order to avoid errors in language analysis. However, when a part of the message is variable, it is difficult to specify the reading and accent of the variable part in advance. Patent Document 1 describes a technique for dynamically generating an accent by applying an accent rule held in advance by the system to a variable part.
Japanese Patent No. 3171775

  However, in Patent Document 1, there is a problem in that a user who describes text cannot specify a rule to be applied to the variable part, and thus a desired synthesized speech may not be generated.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a speech synthesizer capable of processing a text describing a rule applied to a variable part.

  To achieve the above object, a speech synthesizer according to the present invention includes an acquisition unit that acquires a character string to be substituted into a variable part of a text to be synthesized, an extraction processing unit that extracts a rule identifier from the text, Extraction means for extracting information related to the rule from the character string acquired by the acquisition means, a rule to be applied is selected from the rule set and the rule identifier, and at least based on the information related to the rule according to the selected rule And processing means for processing the variable part.

  In addition, a speech synthesizer according to the present invention for achieving the above object includes an acquisition unit that acquires a character string to be substituted into a variable part of a text to be synthesized, and an extraction processing unit that extracts a rule from the text. The image processing apparatus includes: extraction means for extracting information related to the rule from the character string acquired by the acquisition means; and processing means for processing at least the variable part according to the information related to the rule and the rule.

  According to the present invention, a user who describes a text can specify a rule to be applied to the variable part, and thus it is possible to easily generate a synthesized speech desired by the user.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

(First embodiment)
In the first embodiment, a voice synthesizer that provides an answering machine message for an answering machine will be described as an example. With this speech synthesizer, the user can set a desired character string (usually his name) in an answering message. Here, the character string data input by the user is a kana level reading symbol string with an accent symbol.

  The developer of the answering machine creates an answering message template (see FIG. 6A). A template described by a speech synthesis symbol is composed of a combination of three parts: a fixed part, a variable part, and a rule. The fixed part is a fixed message part of the answering message, the variable part is an arbitrary character string input by the user, and the rule is an identification number of a rule applied to the character string.

  In the present embodiment, a pose generation rule inserted before and after the variable part will be described as an example.

  FIG. 4 is an example of a pose generation rule. Rule 1 is a suitable rule when priority is given to ease of listening. Rule 2 is a suitable rule when priority is given to the naturalness of the answering machine message. Rule 3 is a suitable rule when priority is given to information transmission. The pose rule set 201 in FIG. 2 stores these three types of pose generation rules, and the speech synthesizer can use these pose generation rules properly.

  Therefore, the answering machine developer can select (describe in the template) an optimal pose generation rule according to the purchase target. For example, when developing an answering machine for elderly people, it is preferable to select rule 1. In addition, it can be said that rule 2 is suitable for a general answering machine.

  FIG. 1 is a block diagram showing a hardware configuration of the speech synthesizer in the first embodiment.

  In FIG. 1, reference numeral 101 denotes a control memory, which stores the speech synthesis processing procedure of this embodiment and necessary fixed data. A central processing unit 102 performs processing such as numerical calculation / control. Reference numeral 103 denotes a memory in which temporary data is stored. Reference numeral 104 denotes an external storage device in which texts and rules are stored. Reference numeral 105 denotes an input device, which is used by the user to input data to the device and to instruct operations. An output device 106 presents various information to the user under the control of the central processing unit 102. Reference numeral 107 denotes an audio output device that outputs the synthesized content. Reference numeral 108 denotes a bus, and data exchange between the devices is performed through this bus.

  FIG. 2 is a block diagram showing a module configuration of the speech synthesizer in the first embodiment.

  In FIG. 2, the pose rule set 201 stores pose generation rules as shown in FIG. The text holding unit 202 stores an answering message template as shown in FIG. Text generated according to this template is a speech synthesis target. The input processing unit 203 accepts a desired character string input by the user. The character string holding unit 204 holds a character string input by the user. The extraction processing unit 205 extracts a rule number that is an identifier for identifying a rule from the template held by the text holding unit 202. The rule number holding unit 206 holds the extracted rule number. The mora number analysis unit 207 counts the number of mora of the character string held by the character string holding unit 204. The mora number holding unit 208 holds the mora number obtained by the mora number analysis unit 207. The pose generation unit 209 determines a rule to be applied from the pose generation rule and the rule number, and generates poses before and after the variable unit according to the number of mora. The prosody processing unit 210 generates a prosody from the result of the pose generation, the text, and the character string. The waveform generation unit 211 generates synthesized speech based on the prosodic information. The voice output unit 212 outputs synthesized voice.

  FIG. 3 is a flowchart showing the flow of processing of the speech synthesizer in the first embodiment. A control program for executing the flowchart is stored in the control memory 101, the memory 103, the external storage device 104, and the like.

  In step S301, the input processing unit 203 detects a user input. It stays at step S301 until the user inputs. If a user input is detected, the user input is acquired and held in the character string holding unit 204, and the process proceeds to step S302. Since the character string input by the user corresponds to a variable used for the variable part of the template, it will be referred to as a variable in the following description.

  In step S302, the extraction processing unit 205 extracts the rule number from the template held in the text holding unit 202, holds the rule number in the rule number holding unit 206, and then proceeds to step S303.

  In step S303, the mora number analysis unit 207 obtains the mora number of the variable, holds it in the mora number holding unit 208, and proceeds to step S304.

  In step S304, the pose generation unit 209 selects a pose generation rule to be applied from the rule numbers held by the pose rule set 201 and the rule number holding unit 206. Furthermore, after generating poses to be inserted on both sides of a variable (that is, variable part) based on the number of mora held by the mora number holding unit 208, the process proceeds to step S305.

  In step S305, the prosody generation unit 210 generates prosody information from the variables held by the character string holding unit 204, the templates held by the text holding unit 202, and the pose information generated by the pose generation unit 209, and the process proceeds to step S306. .

  In step S306, the waveform generation unit 211 generates synthesized speech corresponding to the text in which the character string held by the character string holding unit 204 is substituted with the template based on the prosodic information, and the process proceeds to step S307. In step S307, the voice output unit 212 outputs the synthesized voice and the process ends.

  FIG. 5 is a diagram showing a pose generation result by using a speech synthesis symbol when rule 1 of the pose generation rule shown in FIG. 4 is applied. FIG. 5A shows an answering message template. FIG. 5B shows a pose generation result when the variable (that is, the character string input by the user) is “tanaka”. Since the number of mora of the variable is 4 mora or less, a 100 ms pause is inserted at both ends of the variable part. FIGS. 5C and 5D show pose generation results when the variables are “Tanakasa 'Toshi” and “Tanaka Kabushikiga' Isha”, respectively. Since both of the variable mora numbers are 5 or more, a pause of 200 ms is inserted in front of the variable portion and 300 ms in the back.

  On the other hand, FIG. 6 is a diagram showing a pose generation result by using a speech synthesis symbol when the rule 2 of the pose generation rule shown in FIG. 4 is applied. FIG. 6A shows an answering machine message template. FIG. 6B shows a pose generation result when the variable is “Tanaka”. Since the number of mora of the variable is 4 or less, no pause is inserted at both ends of the variable part. FIG. 6C shows a pose generation result when the variable is “Tanakasa 'Toshi”. Since the number of mora of the variable is 5 or more, a 100 ms pause is inserted before the variable part. Further, FIG. 6D shows a pose generation result when the variable is “Tanaka Kabushikiga'Isha”. Since the number of mora of the variable is 8 or more, a 100 ms pause is inserted before and after the variable part.

  As described above, by providing a speech synthesizer capable of processing text describing the control rules of the variable part, the answering machine developer can describe the identification number of the pause generation rule in the answering message template. Can do. For this reason, without changing the algorithm of speech synthesis itself, the answering machine corresponding to the purchase target can be developed with a simple process of rewriting the template description. In other words, when developing a certain device, it is possible to provide a framework that allows the device developer to select speech synthesis control suitable for the device.

(Second Embodiment)
In the first embodiment, the case where the control rule acts only on the variable portion has been described. However, the present invention is not limited to this, and the control rule may be configured to act on a portion including the fixed portion.

  In the second embodiment, a voice synthesizer that provides an answering machine message for an answering machine will be described as in the first embodiment.

  FIG. 9 is a diagram showing an example of an accent combining rule for attached words. Rule 1 is a rule that the accent type when the front word and the back word are connected follows the accent type of the front word. Rules 2 and 3 are rules of different accent types when combined depending on whether the front word is a flat plate type or a nucleated type. In general, the accent combination of an adjunct is realized by applying an appropriate rule according to the adjoining adjunct. When analyzing a so-called kanji-kana mixed sentence without control symbols using a word dictionary, the accent combination rule number is stored as an attribute of the attached word, so that appropriate accent combination can be performed. is there. However, when a speech synthesis symbol string consisting of accent symbols and phonetic notations is input, it does not have a word dictionary, so it cannot be known whether a word is an adjunct word, and which of the accent combining rules should be applied. I don't know. In the present embodiment, a framework for appropriately realizing accent coupling for such an input is provided.

  FIG. 7 is a block diagram showing a module configuration of the speech synthesizer in the second embodiment. In FIG. 7, the text holding unit 202, the input processing unit 203, the character string holding unit 204, the rule number holding unit 206, the waveform generation unit 211, and the voice output unit 212 perform the same processing as in the first embodiment. Symbols are used and explanation is omitted.

  The accent combination rule set 701 stores accent combination rules as shown in FIG. The extraction processing unit 205 extracts a rule number from the template held by the text holding unit 202, and further extracts a fixed unit to which the rule is applied together with the variable unit. The fixed part holding part 705 holds a fixed part related to the rule. The accent analysis unit 702 analyzes the accent of the character string held by the character string holding unit 204. The accent holding unit 703 holds the accent information obtained by the accent analysis unit 702. The accent combining unit 704 determines an applied rule from the accent combining rule and the rule number, and generates a combined accent between the variable unit and the fixed unit held by the fixed unit holding unit 705. The prosody generation unit 210 generates prosody information from the result of accent combination, text, and character string.

  FIG. 8 is a flowchart showing the flow of processing of the speech synthesizer in the second embodiment. In step S801, the input processing unit 203 detects a user input. It stays at step S801 until the user inputs. If a user input is detected, the user input is held in the character string holding unit 204, and the process proceeds to step S802. Since the character string input by the user corresponds to a variable used for the variable part of the template, it will be referred to as a variable in the following description.

  In step S <b> 802, the extraction processing unit 205 extracts the rule number from the template held in the text holding unit 202 and holds it in the rule number holding unit 206. Furthermore, after extracting the fixed part involved in the rule and holding it in the fixed part holding part 705, the process proceeds to step S803. In the case of the present embodiment, the fixed portion involved in the rule is a portion sandwiched between the variable portion in the text and the rule.

  In step S803, the accent analysis unit 702 obtains variable accent information, holds it in the accent holding unit 703, and proceeds to step S804. Accent information required by the accent analysis unit 702 includes an accent type and the number of mora. Furthermore, information regarding whether the accent type is a nucleated type or a flat plate type may be included.

  In step S <b> 804, the accent combining unit 704 determines an accent combining rule to be applied from the rule number held by the accent combining rule set 701 and the rule number holding unit 206. Further, after performing accent coupling between the variable part (that is, variable) and the fixed part from the accent information held by the accent holding part 703 and the fixed part held by the fixed part holding part 705, the process proceeds to step S805.

  In step S805, the prosody generation unit 210 generates prosody information from the variable held by the character string holding unit 204, the template held by the text holding unit 202, and the combined accent information generated by the accent combining unit 704, and the process proceeds to step S806. Move.

  In step S806, the waveform generation unit 211 generates synthesized speech based on the prosodic information, and proceeds to step S807. In step S807, the voice output unit 212 outputs the synthesized voice and the process ends.

  FIG. 10 is a diagram showing the result of accent combination when the rule 2 of the accent combination rule shown in FIG. 9 is applied, using symbols for speech synthesis. FIG. 10A shows an answering machine message template. FIG. 10B shows an accent combination result when the variable (that is, the character string input by the user) is “Yamada”. Since the variable accent is flat, the position of the accent kernel is determined according to the accent type of the trailing word “des”. On the other hand, FIG. 10C shows an accent combination result when the variable is “comment”. Since the accent of the variable is nucleated, the position of the accent nucleus after the combination inherits the position of the accent nucleus of the variable as it is.

  As described above, in the second embodiment, by providing a speech synthesizer that processes text describing an accent combining rule, it is possible to appropriately control the combining of accents including not only variable parts but also fixed parts. Is possible.

(Third embodiment)
In the first embodiment and the second embodiment, the case where rules are stored in advance as in the pose rule set 201 or the accent combination rule set 701 has been described. However, the present invention is not limited to this, and the rule itself is a text. It is also good to describe in.

  In the third embodiment as well, as in the second embodiment, an accent combination rule will be described as an example.

  FIG. 11 is a block diagram showing a module configuration of the speech synthesizer in the third embodiment.

  The difference between FIG. 11 and FIG. 7 is two points, that is, there is no accent combination rule set 701 and the rule number holding unit 206 is changed to the rule holding unit 1101.

  The extraction processing unit 205 extracts an accent combining rule from the template held by the text holding unit 202, and further extracts a fixed unit to which the accent combining rule is applied together with the variable unit. The rule holding unit 1101 holds the accent combination rule extracted from the text by the extraction processing unit 205. The accent coupling unit 704 generates a coupling accent between the variable unit and the fixed unit held by the fixed unit holding unit 705 according to the accent coupling rule. The prosody generation unit 210 generates prosody information from the result of accent combination, text, and character string.

  FIG. 12 is a flowchart showing the flow of processing of the speech synthesizer in the third embodiment. In step S1201, the input processing unit 203 detects a user input. It stays at step S1201 until the user inputs. If a user input is detected, the user input is held in the character string holding unit 204, and the process proceeds to step S1202. Since the character string input by the user corresponds to a variable used for the variable part of the template, it will be referred to as a variable in the following description.

  In step S 1202, the extraction processing unit 205 extracts an accent combination rule from the template held in the text holding unit 202 and holds it in the rule number holding unit 206. Furthermore, after extracting the fixed part related to the accent combination rule and holding it in the fixed part holding part 705, the process proceeds to step S1203. In the case of the present embodiment, the fixed part involved in the accent combining rule is a part sandwiched between the variable part in the text and the accent combining rule.

  In step S1203, the accent analysis unit 702 obtains variable accent information, holds it in the accent holding unit 703, and proceeds to step S1204. Accent information required by the accent analysis unit 702 includes an accent type and the number of mora. Furthermore, information regarding whether the accent type is a nucleated type or a flat plate type may be included.

  In step S <b> 1204, the accent combining unit 704 determines from the accent information held by the accent holding unit 703 and the fixed unit held by the fixed unit holding unit 705 according to the accent combining rule held by the rule holding unit 1101. After performing the accent connection between the variable) and the fixed part, the process proceeds to step S1205.

  In step S1205, the prosody generation unit 210 generates prosody information from the variable held by the character string holding unit 204, the template held by the text holding unit 202, and the combined accent information generated by the accent combining unit 704, and the process proceeds to step S1206. Move.

  In step S1206, the waveform generation unit 211 generates synthesized speech based on the prosodic information, and the process proceeds to step S1207. In step S1207, the voice output unit 212 outputs the synthesized voice and the process ends.

  FIG. 13 is an example in which rule 2 of the accent combination rule in FIG. 9 is described directly in the text.

  As described above, in the third embodiment, by providing a speech synthesizer that processes text that directly describes an accent combining rule, it is possible to combine accents without being limited to the accent combining rules prepared in advance. It becomes possible to control appropriately.

(Other embodiments)
In the first to third embodiments, the processing of the speech synthesizer when the rules relating to the pose generation and the accent combination of the appendix are described in the text has been described, but the present invention is not limited to this. Rules regarding reading, other accent generation, prosody generation, and waveform generation may be described.

  In the first embodiment, the example in which the description format of the rule is followed by the variable part has been described. However, the present invention is not limited to this, and as long as the relationship with the variable part is defined, the rule may be placed at any position. It shall be good.

  In 2nd Embodiment and 3rd Embodiment, although the case where the fixing | fixed part which concerns on a rule was arrange | positioned between the variable part and the rule was demonstrated, this invention is not limited to this. As long as the relationship between the fixed part, the variable part, and the rule involved in the rule is defined, it may be arranged at any position.

  Furthermore, the description format of the text is not limited to the format described in the embodiment, and may be any format as long as it can be analyzed.

  The object of the present invention can also be achieved as follows. That is, a storage medium in which a program code of software that realizes the functions of the above-described embodiments is recorded is supplied to the system or apparatus. Then, the computer (or CPU or MPU) of the system or apparatus reads and executes the program code stored in the storage medium. It goes without saying that the purpose is achieved even in this way.

  In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention.

  As a storage medium for supplying the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

  Further, the embodiment according to the present invention is not limited to the case where the functions of the above-described embodiment are realized by executing the program code read by the computer. For example, an OS (operating system) running on a computer performs part or all of actual processing based on an instruction of the program code, and the functions of the above-described embodiments may be realized by the processing. Needless to say, it is included.

  Furthermore, the functions of the embodiment according to the present invention are also realized as follows. That is, the program code read from the storage medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer. Then, based on the instruction of the program code, the CPU provided in the function expansion board or function expansion unit performs part or all of the actual processing. It goes without saying that the functions of the above-described embodiments are realized by this processing.

It is a block diagram which shows the hardware constitutions of the speech synthesizer in 1st Embodiment. It is a block diagram which shows the module structure of the speech synthesizer in 1st Embodiment. It is a flowchart which shows the flow of a process of the speech synthesizer in 1st Embodiment. It is the figure which showed the example of the pose production | generation rule. It is the figure which showed the pose production | generation result at the time of applying the rule 1 of a pose production | generation rule. It is the figure which showed the pose production | generation result at the time of applying the rule 2 of a pose production | generation rule. It is a block diagram which shows the module structure of the speech synthesizer in 2nd Embodiment. It is a flowchart which shows the flow of a process of the speech synthesizer in 2nd Embodiment. It is the figure which showed the example of the accent joint rule. It is the figure which showed the accent joint result at the time of applying the rule 2 of an accent joint rule. It is a block diagram which shows the module structure of the speech synthesizer in 3rd Embodiment. It is a flowchart which shows the flow of a process of the speech synthesizer in 3rd Embodiment. It is a figure which shows an example of the text at the time of describing the accent joint rule directly in the text.

Explanation of symbols

101 Control memory (ROM)
102 Central processing unit 103 Memory (RAM)
104 External Storage Device 105 Input Device 106 Output Device 107 Audio Output Device 108 Bus 201 Pause Rule Set 202 Text Holding Unit 203 Input Processing Unit 204 Character String Holding Unit 205 Extraction Processing Unit 206 Rule Number Holding Unit 207 Mora Number Analysis Unit 208 Number of Mora Holding unit 209 Pose generating unit 210 Prosody generating unit 211 Waveform generating unit 212 Audio output unit 701 Accent combination rule set 702 Accent analysis unit 703 Accent holding unit 704 Accent combining unit 705 Fixed unit holding unit 1101 Rule holding unit

Claims (7)

Acquisition means for acquiring a character string to be substituted into the variable part of the text to be synthesized;
Extraction processing means for extracting a rule identifier from the text;
Extraction means for extracting information related to the rule from the character string acquired by the acquisition means;
Processing means for selecting a rule to be applied from the rule set and the rule identifier, and processing at least a variable part based on information related to the rule according to the selected rule;
A speech synthesizer comprising: The extraction processing means extracts the rule identifier and a fixed part involved in the rule,
The processing means processes the variable part and the fixed part involved in the rule according to information related to the rule and the selected rule.
The speech synthesizer according to claim 1. Acquisition means for acquiring a character string to be substituted into the variable part of the text to be synthesized;
Extraction processing means for extracting rules from the text;
Extraction means for extracting information related to the rule from the character string acquired by the acquisition means;
Processing means for processing at least the variable part according to the information related to the rule and the rule;
A speech synthesizer comprising: An acquisition step of acquiring a character string to be substituted into the variable part of the text to be synthesized;
An extraction processing step for extracting a rule identifier from the text;
An extraction step of extracting information related to the rule from the character string acquired in the acquisition step;
A processing step of selecting a rule to be applied from a rule set for speech synthesis and the rule identifier, and processing at least a variable part based on information related to the rule according to the selected rule;
A speech synthesis method comprising: The extraction processing step extracts the fixed part involved in the rule identifier and the rule,
The processing step processes the variable part and the fixed part involved in the rule according to the information related to the rule and the selected rule.
The speech synthesis method according to claim 4. An acquisition step of acquiring a character string to be substituted into the variable part of the text to be synthesized;
An extraction processing step for extracting rules from the text;
An extraction step of extracting information related to the rule from the character string acquired in the acquisition step;
Processing steps for processing at least the variable part according to the information related to the rule and the rule;
A speech synthesis method comprising: A control program for causing a computer to execute the speech synthesis method according to claim 4.

Images